diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample82-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample82-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..680180d8d0b3a2dc6cdcadeaac197941aee1d5c9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample82-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eac5ce4be8860ddb45285d5affd2772d0f687cba21f593120b657f04fc6b7a38 +size 2874781 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample84-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample84-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7f8374363822163fd64a89707d7de2e4425f1961 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample84-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e25b8ceff385e051ace6ddcadecdbd2423f7ce8b8be1d4978b34f406a853153 +size 3039503 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample85-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample85-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..8241d5664a907dfadded2b79f4b08cc24891cf57 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample85-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4dcb7840aeba4e17cca50e47c933ee20921437405b12ecda5c9a215493f808be +size 2996525 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample86-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample86-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..87ce0785114c7958bb3accc7eed769da7a1770b1 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample86-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14d905ca0a1ce782856f636f86ff77d147da18113355d7824d81883e4b65461e +size 2872189 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample87-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample87-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..aff2753aecd96e9b53343e1970cc922ca2ddc921 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample87-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad354352d89ff4b3ff76aeb0dfd09792fe823adbb6f4147f68dcd2f36a6bb63e +size 3091317 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample88-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample88-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..167684595d6d7204cbc7b71ea59e19131e95ed3b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample88-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da906860e9756f61a473283f91aace55672e61af06a5a515bcebb8dcb4b26834 +size 3030446 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample89-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample89-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..a1a104e0d2e08187109bd5936546defa3e21c28c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample89-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9a0503abbc311db70c5668314669bb7347e742d127258192b4b55ae32f83f91 +size 2968073 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample9-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample9-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7d15497c6c239e7b9b4f1f71f8efeb2fb358716c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample9-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:266a50c956040432ff2ef945794e0b7b1e521e567f33aeefa4ec3b639f7f8637 +size 3812859 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample90-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample90-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..48937252e2cb5698fb2f3373548141a90c192497 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample90-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d57be96c65964eed32d7ec139caa838ced55520c4d428e7fc4f51cbd8575d25 +size 3055844 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample91-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample91-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..cdbaf49a8a9bcaa8a1d036b754b371ae2baf7bc5 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample91-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:797c313527f7df6bda8bc0fec8a21e3beaa6ccb3f1bcf8814ab4f8bc743b2548 +size 2965306 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample92-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample92-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d22a00586bcdd43a616137e7259c498a06c90579 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample92-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:84e37f4bf2c65e9dd8098e45917628460c96e264b68d7538a4033feb98b78cc8 +size 3012374 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample94-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample94-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..59cc151c6a2b8f6ae6166ac4608b1d1ec1860115 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample94-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:adb9c4ae5dea9dfed5f9ea3038dbe24e145bb3082e3fd74b60ae9bbd45a4f566 +size 3002636 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample96-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample96-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..14b16fafd7632b73202f519dee200e7b08a48281 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample96-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec2da7b8fdfa674e969bc965b6683ec157cd3cc3903a857febfff364e096fc93 +size 2748902 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample97-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample97-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d2a08e6589a25db5b96fab999c87912637292f98 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample97-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd181be1d819894ef397dfc02918d3653a90fc8c8425e4b259d70eeac805c30e +size 2860436 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample98-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample98-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..bd837bbc1f2f7df0278e11c024da163fbfdcbaac --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample98-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e20c4f19c39377df09a378e78bce18ae43ddd1f8e7135fc725ee44509557351 +size 2772272 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample99-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample99-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..04fdce1b498733cc182f54518e1361a3dfdbc032 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_arc_challenge/sample99-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57023d21aaff43342fa8c0aa8c57d6a207aae9f28d538bc63994344915f06f60 +size 2899145 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/dtufc_hyperprior-featurecoding_qwen_individual.log b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/dtufc_hyperprior-featurecoding_qwen_individual.log index 557293adc2f3e6dc171191de0dad1e9abc1a2b3f..163fa2d2f7735432565e394fb5e56096689be149 100644 --- a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/dtufc_hyperprior-featurecoding_qwen_individual.log +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/dtufc_hyperprior-featurecoding_qwen_individual.log @@ -1,16958 +1,3 @@ -Experiment: dtufc_hyperprior-featurecoding_qwen_individual -Log file: output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/dtufc_hyperprior-featurecoding_qwen_individual.log -DTUFCCodecConfig: - arch: hyperprior-featurecoding - handler: qwen - checkpoint: codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar - transform_type: kmeans - transform_mapping:featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json - bit_depth: 8 - device: cuda:0 -Loading checkpoint: codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar -Checkpoint epoch: 599 -Loaded hyperprior-featurecoding (1-channel) on cuda:0 -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.0.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.0.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.1.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.1.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.2.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.2.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.3.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.3.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.output' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json -Loaded per-key mappings: model=qwen - Keys: ['layer.0.k_cache', 'layer.0.v_cache', 'layer.1.k_cache', 'layer.1.v_cache', 'layer.2.k_cache', 'layer.2.v_cache', 'layer.3.k_cache', 'layer.3.v_cache', 'layer.4.k_cache', 'layer.4.v_cache', 'layer.4.output'] ----------------- ----------------------------------------------------------------------------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding -Checkpoint codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar -Transform type kmeans -Transform config featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json -Input ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k -Output output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k ----------------- ----------------------------------------------------------------------------------------------------------------------------- -Files found: 100 ----------------------------------------------------------------------- - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1-layer4-item1.zst (1/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.010s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 173, 128) -Output shape: (1, 173, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) - layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 16,468B, BPFP=0.7437 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 42,400B, BPFP=1.9147 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 29,668B, BPFP=1.3398 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 45,136B, BPFP=2.0383 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 33,332B, BPFP=1.5052 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 45,944B, BPFP=2.0748 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 34,432B, BPFP=1.5549 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 45,264B, BPFP=2.0441 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 30,184B, BPFP=1.3631 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 46,332B, BPFP=2.0923 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 130,608B, BPFP=1.4745 -⌛️ [2/4] FRONTEND: Frontend time: 0.490s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 173, 128]) - layer.0.v_cache: torch.Size([1, 8, 173, 128]) - layer.1.k_cache: torch.Size([1, 8, 173, 128]) - layer.1.v_cache: torch.Size([1, 8, 173, 128]) - layer.2.k_cache: torch.Size([1, 8, 173, 128]) - layer.2.v_cache: torch.Size([1, 8, 173, 128]) - layer.3.k_cache: torch.Size([1, 8, 173, 128]) - layer.3.v_cache: torch.Size([1, 8, 173, 128]) - layer.4.k_cache: torch.Size([1, 8, 173, 128]) - layer.4.v_cache: torch.Size([1, 8, 173, 128]) - layer.4.output: torch.Size([1, 173, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.444s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 173, 128]) - layer.0.v_cache: torch.Size([1, 8, 173, 128]) - layer.1.k_cache: torch.Size([1, 8, 173, 128]) - layer.1.v_cache: torch.Size([1, 8, 173, 128]) - layer.2.k_cache: torch.Size([1, 8, 173, 128]) - layer.2.v_cache: torch.Size([1, 8, 173, 128]) - layer.3.k_cache: torch.Size([1, 8, 173, 128]) - layer.3.v_cache: torch.Size([1, 8, 173, 128]) - layer.4.k_cache: torch.Size([1, 8, 173, 128]) - layer.4.v_cache: torch.Size([1, 8, 173, 128]) - layer.4.output: torch.Size([1, 173, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02440321 7.49046088 - layer.0.v_cache 0.00000027 0.00022737 - layer.1.k_cache 0.00314874 0.75185725 - layer.1.v_cache 0.00000072 0.00077331 - layer.2.k_cache 0.00114712 0.39501984 - layer.2.v_cache 0.00000110 0.00114576 - layer.3.k_cache 0.00139354 0.45736408 - layer.3.v_cache 0.00000204 0.00188292 - layer.4.k_cache 0.00353492 0.88896832 - layer.4.v_cache 0.00000301 0.00315314 - layer.4.output 0.00018661 0.06732985 - ------------------------------------------------------------------------------------- - TOTAL 0.00245579 0.73286945 - (elements=2,480,128) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2480128 -Total Bytes 499768 -BPFP 1.6121 bits/point -EBPFP 3.2241 equivalent bits/point -MSE 0.732869 ----------------------- -------------------------------------------------------- -Time: 0.944s Load: 0.010s, Pack+Encode: 0.490s, Decode+Unpack: 0.444s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7329 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample100-layer4-item1.zst (2/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample100-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 109, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 109, 128) -Output shape: (1, 109, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.output: torch.Size([1, 109, 4096]) -> torch.Size([1, 1, 109, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,136B, BPFP=0.7265 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 27,560B, BPFP=1.9753 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,516B, BPFP=1.3988 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 29,784B, BPFP=2.1347 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,888B, BPFP=1.5688 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,024B, BPFP=2.1519 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,776B, BPFP=1.6325 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 29,548B, BPFP=2.1178 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 19,828B, BPFP=1.4212 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,252B, BPFP=2.1683 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 84,220B, BPFP=1.5091 -⌛️ [2/4] FRONTEND: Frontend time: 0.270s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.output: torch.Size([1, 109, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.310s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.output: torch.Size([1, 109, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02465495 8.05804443 - layer.0.v_cache 0.00000027 0.00022998 - layer.1.k_cache 0.00330346 0.82048266 - layer.1.v_cache 0.00000080 0.00080155 - layer.2.k_cache 0.00115767 0.39721232 - layer.2.v_cache 0.00000114 0.00110043 - layer.3.k_cache 0.00132842 0.46342741 - layer.3.v_cache 0.00000211 0.00180431 - layer.4.k_cache 0.00335301 0.86559282 - layer.4.v_cache 0.00000290 0.00295481 - layer.4.output 0.00024947 0.07567812 - ------------------------------------------------------------------------------------- - TOTAL 0.00248590 0.77959737 - (elements=1,562,624) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1562624 -Total Bytes 325532 -BPFP 1.6666 bits/point -EBPFP 3.3332 equivalent bits/point -MSE 0.779597 ----------------------- -------------------------------------------------------- -Time: 0.585s Load: 0.006s, Pack+Encode: 0.270s, Decode+Unpack: 0.310s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 109, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7796 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample100-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample100-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample108-layer4-item1.zst (3/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample108-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 97, 128) -Output shape: (1, 97, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.output: torch.Size([1, 97, 4096]) -> torch.Size([1, 1, 97, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,448B, BPFP=0.7610 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 24,172B, BPFP=1.9468 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,696B, BPFP=1.3447 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,804B, BPFP=2.0783 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,160B, BPFP=1.5432 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,372B, BPFP=2.1240 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,912B, BPFP=1.6037 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,832B, BPFP=2.0805 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,260B, BPFP=1.3901 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,540B, BPFP=2.1376 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 74,736B, BPFP=1.5048 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02579871 8.36424208 - layer.0.v_cache 0.00000028 0.00024121 - layer.1.k_cache 0.00331373 0.80556417 - layer.1.v_cache 0.00000086 0.00089134 - layer.2.k_cache 0.00113316 0.43120048 - layer.2.v_cache 0.00000115 0.00120392 - layer.3.k_cache 0.00137718 0.47971022 - layer.3.v_cache 0.00000224 0.00202165 - layer.4.k_cache 0.00334605 0.88578167 - layer.4.v_cache 0.00000324 0.00346960 - layer.4.output 0.00031880 0.08460554 - ------------------------------------------------------------------------------------- - TOTAL 0.00258942 0.80805346 - (elements=1,390,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1390592 -Total Bytes 285932 -BPFP 1.6450 bits/point -EBPFP 3.2899 equivalent bits/point -MSE 0.808053 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.008s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8081 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample108-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample108-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1087-layer4-item1.zst (4/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1087-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 50, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.003s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 50, 128) -Output shape: (1, 50, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.0.v_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.1.k_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.1.v_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.2.k_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.2.v_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.3.k_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.3.v_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.4.k_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.4.v_cache: torch.Size([1, 8, 50, 128]) -> torch.Size([1, 1, 50, 1024]) - layer.4.output: torch.Size([1, 50, 4096]) -> torch.Size([1, 1, 50, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 5,376B, BPFP=0.8400 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 13,148B, BPFP=2.0544 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 9,572B, BPFP=1.4956 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 14,224B, BPFP=2.2225 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 11,044B, BPFP=1.7256 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 14,804B, BPFP=2.3131 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 11,304B, BPFP=1.7663 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 14,472B, BPFP=2.2612 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 9,740B, BPFP=1.5219 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 14,640B, BPFP=2.2875 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 42,020B, BPFP=1.6414 -⌛️ [2/4] FRONTEND: Frontend time: 0.219s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 50, 128]) - layer.0.v_cache: torch.Size([1, 8, 50, 128]) - layer.1.k_cache: torch.Size([1, 8, 50, 128]) - layer.1.v_cache: torch.Size([1, 8, 50, 128]) - layer.2.k_cache: torch.Size([1, 8, 50, 128]) - layer.2.v_cache: torch.Size([1, 8, 50, 128]) - layer.3.k_cache: torch.Size([1, 8, 50, 128]) - layer.3.v_cache: torch.Size([1, 8, 50, 128]) - layer.4.k_cache: torch.Size([1, 8, 50, 128]) - layer.4.v_cache: torch.Size([1, 8, 50, 128]) - layer.4.output: torch.Size([1, 50, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.213s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 50, 128]) - layer.0.v_cache: torch.Size([1, 8, 50, 128]) - layer.1.k_cache: torch.Size([1, 8, 50, 128]) - layer.1.v_cache: torch.Size([1, 8, 50, 128]) - layer.2.k_cache: torch.Size([1, 8, 50, 128]) - layer.2.v_cache: torch.Size([1, 8, 50, 128]) - layer.3.k_cache: torch.Size([1, 8, 50, 128]) - layer.3.v_cache: torch.Size([1, 8, 50, 128]) - layer.4.k_cache: torch.Size([1, 8, 50, 128]) - layer.4.v_cache: torch.Size([1, 8, 50, 128]) - layer.4.output: torch.Size([1, 50, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02911560 11.71019409 - layer.0.v_cache 0.00000029 0.00024933 - layer.1.k_cache 0.00403332 1.04451477 - layer.1.v_cache 0.00000087 0.00088725 - layer.2.k_cache 0.00117108 0.47603874 - layer.2.v_cache 0.00000108 0.00118187 - layer.3.k_cache 0.00141424 0.54856674 - layer.3.v_cache 0.00000205 0.00191995 - layer.4.k_cache 0.00319313 1.02655006 - layer.4.v_cache 0.00000283 0.00313498 - layer.4.output 0.00027592 0.13221746 - ------------------------------------------------------------------------------------- - TOTAL 0.00285987 1.09586483 - (elements=716,800) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 716800 -Total Bytes 160344 -BPFP 1.7896 bits/point -EBPFP 3.5791 equivalent bits/point -MSE 1.095865 ----------------------- -------------------------------------------------------- -Time: 0.435s Load: 0.003s, Pack+Encode: 0.219s, Decode+Unpack: 0.213s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 50, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 50, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.0959 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1087-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1087-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1128-layer4-item1.zst (5/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1128-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 57, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 57, 128) -Output shape: (1, 57, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.0.v_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.1.k_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.1.v_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.2.k_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.2.v_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.3.k_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.3.v_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.4.k_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.4.v_cache: torch.Size([1, 8, 57, 128]) -> torch.Size([1, 1, 57, 1024]) - layer.4.output: torch.Size([1, 57, 4096]) -> torch.Size([1, 1, 57, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 5,716B, BPFP=0.7834 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 14,004B, BPFP=1.9194 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 10,460B, BPFP=1.4337 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 15,480B, BPFP=2.1217 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 11,660B, BPFP=1.5981 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 16,204B, BPFP=2.2209 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 12,172B, BPFP=1.6683 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 15,900B, BPFP=2.1793 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 10,536B, BPFP=1.4441 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 16,048B, BPFP=2.1996 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 45,608B, BPFP=1.5628 -⌛️ [2/4] FRONTEND: Frontend time: 0.157s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 57, 128]) - layer.0.v_cache: torch.Size([1, 8, 57, 128]) - layer.1.k_cache: torch.Size([1, 8, 57, 128]) - layer.1.v_cache: torch.Size([1, 8, 57, 128]) - layer.2.k_cache: torch.Size([1, 8, 57, 128]) - layer.2.v_cache: torch.Size([1, 8, 57, 128]) - layer.3.k_cache: torch.Size([1, 8, 57, 128]) - layer.3.v_cache: torch.Size([1, 8, 57, 128]) - layer.4.k_cache: torch.Size([1, 8, 57, 128]) - layer.4.v_cache: torch.Size([1, 8, 57, 128]) - layer.4.output: torch.Size([1, 57, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.209s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 57, 128]) - layer.0.v_cache: torch.Size([1, 8, 57, 128]) - layer.1.k_cache: torch.Size([1, 8, 57, 128]) - layer.1.v_cache: torch.Size([1, 8, 57, 128]) - layer.2.k_cache: torch.Size([1, 8, 57, 128]) - layer.2.v_cache: torch.Size([1, 8, 57, 128]) - layer.3.k_cache: torch.Size([1, 8, 57, 128]) - layer.3.v_cache: torch.Size([1, 8, 57, 128]) - layer.4.k_cache: torch.Size([1, 8, 57, 128]) - layer.4.v_cache: torch.Size([1, 8, 57, 128]) - layer.4.output: torch.Size([1, 57, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02901734 10.43189547 - layer.0.v_cache 0.00000028 0.00023632 - layer.1.k_cache 0.00371865 0.91275948 - layer.1.v_cache 0.00000081 0.00082479 - layer.2.k_cache 0.00114644 0.42774823 - layer.2.v_cache 0.00000109 0.00114488 - layer.3.k_cache 0.00141512 0.50165176 - layer.3.v_cache 0.00000208 0.00182851 - layer.4.k_cache 0.00325023 0.93998022 - layer.4.v_cache 0.00000285 0.00299744 - layer.4.output 0.00023785 0.10173645 - ------------------------------------------------------------------------------------- - TOTAL 0.00282188 0.97342949 - (elements=817,152) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 817152 -Total Bytes 173788 -BPFP 1.7014 bits/point -EBPFP 3.4028 equivalent bits/point -MSE 0.973429 ----------------------- -------------------------------------------------------- -Time: 0.371s Load: 0.005s, Pack+Encode: 0.157s, Decode+Unpack: 0.209s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 57, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 57, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9734 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1128-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1128-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample117-layer4-item1.zst (6/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample117-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 113, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 113, 128) -Output shape: (1, 113, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.output: torch.Size([1, 113, 4096]) -> torch.Size([1, 1, 113, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,784B, BPFP=0.7456 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,536B, BPFP=1.9729 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,968B, BPFP=1.3805 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,536B, BPFP=2.1112 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,548B, BPFP=1.5589 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,076B, BPFP=2.1485 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,304B, BPFP=1.6112 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,580B, BPFP=2.1142 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,108B, BPFP=1.3902 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,224B, BPFP=2.1587 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 90,192B, BPFP=1.5589 -⌛️ [2/4] FRONTEND: Frontend time: 0.214s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.output: torch.Size([1, 113, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.output: torch.Size([1, 113, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.03016301 9.23314782 - layer.0.v_cache 0.00000028 0.00023785 - layer.1.k_cache 0.00339028 0.77339767 - layer.1.v_cache 0.00000081 0.00080444 - layer.2.k_cache 0.00113693 0.40245171 - layer.2.v_cache 0.00000114 0.00113113 - layer.3.k_cache 0.00135151 0.45612355 - layer.3.v_cache 0.00000219 0.00185921 - layer.4.k_cache 0.00328814 0.79347445 - layer.4.v_cache 0.00000314 0.00321924 - layer.4.output 0.00017637 0.07366753 - ------------------------------------------------------------------------------------- - TOTAL 0.00286021 0.85432266 - (elements=1,619,968) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1619968 -Total Bytes 338856 -BPFP 1.6734 bits/point -EBPFP 3.3468 equivalent bits/point -MSE 0.854323 ----------------------- -------------------------------------------------------- -Time: 0.520s Load: 0.008s, Pack+Encode: 0.214s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 113, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8543 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample117-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample117-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample120-layer4-item1.zst (7/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample120-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 108, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 108, 128) -Output shape: (1, 108, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.0.v_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.1.k_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.1.v_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.2.k_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.2.v_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.3.k_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.3.v_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.4.k_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.4.v_cache: torch.Size([1, 8, 108, 128]) -> torch.Size([1, 1, 108, 1024]) - layer.4.output: torch.Size([1, 108, 4096]) -> torch.Size([1, 1, 108, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,692B, BPFP=0.7011 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 26,884B, BPFP=1.9447 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,204B, BPFP=1.3892 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 29,388B, BPFP=2.1259 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,648B, BPFP=1.5660 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 29,712B, BPFP=2.1493 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,652B, BPFP=1.6386 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 29,360B, BPFP=2.1238 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 19,368B, BPFP=1.4010 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,140B, BPFP=2.1803 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 83,120B, BPFP=1.5032 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 108, 128]) - layer.0.v_cache: torch.Size([1, 8, 108, 128]) - layer.1.k_cache: torch.Size([1, 8, 108, 128]) - layer.1.v_cache: torch.Size([1, 8, 108, 128]) - layer.2.k_cache: torch.Size([1, 8, 108, 128]) - layer.2.v_cache: torch.Size([1, 8, 108, 128]) - layer.3.k_cache: torch.Size([1, 8, 108, 128]) - layer.3.v_cache: torch.Size([1, 8, 108, 128]) - layer.4.k_cache: torch.Size([1, 8, 108, 128]) - layer.4.v_cache: torch.Size([1, 8, 108, 128]) - layer.4.output: torch.Size([1, 108, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 108, 128]) - layer.0.v_cache: torch.Size([1, 8, 108, 128]) - layer.1.k_cache: torch.Size([1, 8, 108, 128]) - layer.1.v_cache: torch.Size([1, 8, 108, 128]) - layer.2.k_cache: torch.Size([1, 8, 108, 128]) - layer.2.v_cache: torch.Size([1, 8, 108, 128]) - layer.3.k_cache: torch.Size([1, 8, 108, 128]) - layer.3.v_cache: torch.Size([1, 8, 108, 128]) - layer.4.k_cache: torch.Size([1, 8, 108, 128]) - layer.4.v_cache: torch.Size([1, 8, 108, 128]) - layer.4.output: torch.Size([1, 108, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02569981 8.68038771 - layer.0.v_cache 0.00000028 0.00022513 - layer.1.k_cache 0.00351470 0.82272770 - layer.1.v_cache 0.00000073 0.00076295 - layer.2.k_cache 0.00116483 0.40615061 - layer.2.v_cache 0.00000104 0.00110065 - layer.3.k_cache 0.00137767 0.46103915 - layer.3.v_cache 0.00000202 0.00179384 - layer.4.k_cache 0.00329212 0.87482693 - layer.4.v_cache 0.00000308 0.00300115 - layer.4.output 0.00021651 0.07640783 - ------------------------------------------------------------------------------------- - TOTAL 0.00256588 0.82554622 - (elements=1,548,288) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1548288 -Total Bytes 321168 -BPFP 1.6595 bits/point -EBPFP 3.3189 equivalent bits/point -MSE 0.825546 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.008s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 108, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 108, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8255 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample120-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample120-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1295-layer4-item1.zst (8/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1295-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 47, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.004s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 47, 128) -Output shape: (1, 47, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.0.v_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.1.k_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.1.v_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.2.k_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.2.v_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.3.k_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.3.v_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.4.k_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.4.v_cache: torch.Size([1, 8, 47, 128]) -> torch.Size([1, 1, 47, 1024]) - layer.4.output: torch.Size([1, 47, 4096]) -> torch.Size([1, 1, 47, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 4,980B, BPFP=0.8278 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 12,568B, BPFP=2.0891 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 9,240B, BPFP=1.5359 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 13,608B, BPFP=2.2620 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 10,352B, BPFP=1.7207 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 14,176B, BPFP=2.3564 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 10,724B, BPFP=1.7826 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 13,960B, BPFP=2.3205 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 9,128B, BPFP=1.5173 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 14,068B, BPFP=2.3384 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 39,896B, BPFP=1.6579 -⌛️ [2/4] FRONTEND: Frontend time: 0.158s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 47, 128]) - layer.0.v_cache: torch.Size([1, 8, 47, 128]) - layer.1.k_cache: torch.Size([1, 8, 47, 128]) - layer.1.v_cache: torch.Size([1, 8, 47, 128]) - layer.2.k_cache: torch.Size([1, 8, 47, 128]) - layer.2.v_cache: torch.Size([1, 8, 47, 128]) - layer.3.k_cache: torch.Size([1, 8, 47, 128]) - layer.3.v_cache: torch.Size([1, 8, 47, 128]) - layer.4.k_cache: torch.Size([1, 8, 47, 128]) - layer.4.v_cache: torch.Size([1, 8, 47, 128]) - layer.4.output: torch.Size([1, 47, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.209s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 47, 128]) - layer.0.v_cache: torch.Size([1, 8, 47, 128]) - layer.1.k_cache: torch.Size([1, 8, 47, 128]) - layer.1.v_cache: torch.Size([1, 8, 47, 128]) - layer.2.k_cache: torch.Size([1, 8, 47, 128]) - layer.2.v_cache: torch.Size([1, 8, 47, 128]) - layer.3.k_cache: torch.Size([1, 8, 47, 128]) - layer.3.v_cache: torch.Size([1, 8, 47, 128]) - layer.4.k_cache: torch.Size([1, 8, 47, 128]) - layer.4.v_cache: torch.Size([1, 8, 47, 128]) - layer.4.output: torch.Size([1, 47, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02987218 12.93321975 - layer.0.v_cache 0.00000031 0.00026311 - layer.1.k_cache 0.00397950 1.11735876 - layer.1.v_cache 0.00000076 0.00083647 - layer.2.k_cache 0.00125400 0.46429768 - layer.2.v_cache 0.00000109 0.00120511 - layer.3.k_cache 0.00146534 0.53310816 - layer.3.v_cache 0.00000206 0.00193578 - layer.4.k_cache 0.00325167 1.05562234 - layer.4.v_cache 0.00000289 0.00317354 - layer.4.output 0.00022000 0.11204733 - ------------------------------------------------------------------------------------- - TOTAL 0.00290784 1.18280071 - (elements=673,792) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 673792 -Total Bytes 152700 -BPFP 1.8130 bits/point -EBPFP 3.6260 equivalent bits/point -MSE 1.182801 ----------------------- -------------------------------------------------------- -Time: 0.371s Load: 0.004s, Pack+Encode: 0.158s, Decode+Unpack: 0.209s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 47, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 47, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.1828 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample1295-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1295-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample13-layer4-item1.zst (9/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample13-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 154, 128) -Output shape: (1, 154, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.0.v_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.1.k_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.1.v_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.2.k_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.2.v_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.3.k_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.3.v_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.4.k_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.4.v_cache: torch.Size([1, 8, 154, 128]) -> torch.Size([1, 1, 154, 1024]) - layer.4.output: torch.Size([1, 154, 4096]) -> torch.Size([1, 1, 154, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 15,184B, BPFP=0.7703 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 38,044B, BPFP=1.9300 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 27,200B, BPFP=1.3799 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 40,792B, BPFP=2.0694 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 30,752B, BPFP=1.5601 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 42,128B, BPFP=2.1372 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 31,968B, BPFP=1.6218 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 41,776B, BPFP=2.1193 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 27,856B, BPFP=1.4131 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 42,648B, BPFP=2.1636 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 123,784B, BPFP=1.5699 -⌛️ [2/4] FRONTEND: Frontend time: 0.262s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 154, 128]) - layer.0.v_cache: torch.Size([1, 8, 154, 128]) - layer.1.k_cache: torch.Size([1, 8, 154, 128]) - layer.1.v_cache: torch.Size([1, 8, 154, 128]) - layer.2.k_cache: torch.Size([1, 8, 154, 128]) - layer.2.v_cache: torch.Size([1, 8, 154, 128]) - layer.3.k_cache: torch.Size([1, 8, 154, 128]) - layer.3.v_cache: torch.Size([1, 8, 154, 128]) - layer.4.k_cache: torch.Size([1, 8, 154, 128]) - layer.4.v_cache: torch.Size([1, 8, 154, 128]) - layer.4.output: torch.Size([1, 154, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.399s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 154, 128]) - layer.0.v_cache: torch.Size([1, 8, 154, 128]) - layer.1.k_cache: torch.Size([1, 8, 154, 128]) - layer.1.v_cache: torch.Size([1, 8, 154, 128]) - layer.2.k_cache: torch.Size([1, 8, 154, 128]) - layer.2.v_cache: torch.Size([1, 8, 154, 128]) - layer.3.k_cache: torch.Size([1, 8, 154, 128]) - layer.3.v_cache: torch.Size([1, 8, 154, 128]) - layer.4.k_cache: torch.Size([1, 8, 154, 128]) - layer.4.v_cache: torch.Size([1, 8, 154, 128]) - layer.4.output: torch.Size([1, 154, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02515895 7.74686977 - layer.0.v_cache 0.00000027 0.00022498 - layer.1.k_cache 0.00328501 0.71356954 - layer.1.v_cache 0.00000091 0.00080404 - layer.2.k_cache 0.00115449 0.40141940 - layer.2.v_cache 0.00000127 0.00117799 - layer.3.k_cache 0.00140159 0.45536745 - layer.3.v_cache 0.00000237 0.00196566 - layer.4.k_cache 0.00336527 0.80911389 - layer.4.v_cache 0.00000311 0.00311979 - layer.4.output 0.00016660 0.07289405 - ------------------------------------------------------------------------------------- - TOTAL 0.00250283 0.74465776 - (elements=2,207,744) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2207744 -Total Bytes 462132 -BPFP 1.6746 bits/point -EBPFP 3.3492 equivalent bits/point -MSE 0.744658 ----------------------- -------------------------------------------------------- -Time: 0.669s Load: 0.008s, Pack+Encode: 0.262s, Decode+Unpack: 0.399s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 154, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7447 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample13-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample13-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample130-layer4-item1.zst (10/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample130-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 100, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 100, 128) -Output shape: (1, 100, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.output: torch.Size([1, 100, 4096]) -> torch.Size([1, 1, 100, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,948B, BPFP=0.7772 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 25,200B, BPFP=1.9688 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,840B, BPFP=1.3938 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 27,228B, BPFP=2.1272 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 20,372B, BPFP=1.5916 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 27,588B, BPFP=2.1553 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,940B, BPFP=1.6359 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 27,384B, BPFP=2.1394 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 18,220B, BPFP=1.4234 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 27,936B, BPFP=2.1825 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 79,824B, BPFP=1.5591 -⌛️ [2/4] FRONTEND: Frontend time: 0.211s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.output: torch.Size([1, 100, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.297s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.output: torch.Size([1, 100, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02583573 9.20972046 - layer.0.v_cache 0.00000027 0.00023822 - layer.1.k_cache 0.00330333 0.90073868 - layer.1.v_cache 0.00000088 0.00082816 - layer.2.k_cache 0.00114776 0.42334045 - layer.2.v_cache 0.00000114 0.00113883 - layer.3.k_cache 0.00139474 0.49695652 - layer.3.v_cache 0.00000229 0.00192720 - layer.4.k_cache 0.00328125 0.90875954 - layer.4.v_cache 0.00000316 0.00320945 - layer.4.output 0.00020624 0.09864681 - ------------------------------------------------------------------------------------- - TOTAL 0.00255682 0.88153177 - (elements=1,433,600) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1433600 -Total Bytes 302480 -BPFP 1.6879 bits/point -EBPFP 3.3759 equivalent bits/point -MSE 0.881532 ----------------------- -------------------------------------------------------- -Time: 0.514s Load: 0.007s, Pack+Encode: 0.211s, Decode+Unpack: 0.297s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 100, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8815 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample130-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample130-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample144-layer4-item1.zst (11/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample144-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 95, 128) -Output shape: (1, 95, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.output: torch.Size([1, 95, 4096]) -> torch.Size([1, 1, 95, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,704B, BPFP=0.7980 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,924B, BPFP=1.9674 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,808B, BPFP=1.3822 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,404B, BPFP=2.0891 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,064B, BPFP=1.5678 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,884B, BPFP=2.1286 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,928B, BPFP=1.6388 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,716B, BPFP=2.1148 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,324B, BPFP=1.4247 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,360B, BPFP=2.1678 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 72,508B, BPFP=1.4907 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02533972 8.89160413 - layer.0.v_cache 0.00000026 0.00023485 - layer.1.k_cache 0.00345234 0.88137400 - layer.1.v_cache 0.00000076 0.00084525 - layer.2.k_cache 0.00126989 0.43317305 - layer.2.v_cache 0.00000106 0.00113651 - layer.3.k_cache 0.00140666 0.48437496 - layer.3.v_cache 0.00000216 0.00198204 - layer.4.k_cache 0.00351815 0.91273177 - layer.4.v_cache 0.00000309 0.00341619 - layer.4.output 0.00017514 0.09156812 - ------------------------------------------------------------------------------------- - TOTAL 0.00254962 0.85551038 - (elements=1,361,920) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1361920 -Total Bytes 282624 -BPFP 1.6602 bits/point -EBPFP 3.3203 equivalent bits/point -MSE 0.855510 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8555 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample144-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample144-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample145-layer4-item1.zst (12/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample145-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 97, 128) -Output shape: (1, 97, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.output: torch.Size([1, 97, 4096]) -> torch.Size([1, 1, 97, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,172B, BPFP=0.7387 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,676B, BPFP=1.9069 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,824B, BPFP=1.3550 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,456B, BPFP=2.0503 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,956B, BPFP=1.5267 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,020B, BPFP=2.0957 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,732B, BPFP=1.5892 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,672B, BPFP=2.0677 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,080B, BPFP=1.3756 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,472B, BPFP=2.1321 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 72,456B, BPFP=1.4589 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02570412 8.09418920 - layer.0.v_cache 0.00000027 0.00023117 - layer.1.k_cache 0.00344234 0.79440669 - layer.1.v_cache 0.00000085 0.00080074 - layer.2.k_cache 0.00116154 0.42324872 - layer.2.v_cache 0.00000108 0.00114047 - layer.3.k_cache 0.00150460 0.48813452 - layer.3.v_cache 0.00000231 0.00188497 - layer.4.k_cache 0.00334272 0.87631414 - layer.4.v_cache 0.00000310 0.00327823 - layer.4.output 0.00020164 0.08367318 - ------------------------------------------------------------------------------------- - TOTAL 0.00256925 0.78702297 - (elements=1,390,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1390592 -Total Bytes 281516 -BPFP 1.6195 bits/point -EBPFP 3.2391 equivalent bits/point -MSE 0.787023 ----------------------- -------------------------------------------------------- -Time: 0.511s Load: 0.006s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7870 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample145-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample145-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample146-layer4-item1.zst (13/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample146-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 98, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 98, 128) -Output shape: (1, 98, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.output: torch.Size([1, 98, 4096]) -> torch.Size([1, 1, 98, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,528B, BPFP=0.7596 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 24,460B, BPFP=1.9499 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,116B, BPFP=1.3645 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 26,320B, BPFP=2.0982 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,428B, BPFP=1.5488 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,900B, BPFP=2.1445 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,020B, BPFP=1.5960 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,580B, BPFP=2.1189 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,464B, BPFP=1.3922 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 27,300B, BPFP=2.1763 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 79,032B, BPFP=1.5751 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.output: torch.Size([1, 98, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.output: torch.Size([1, 98, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02660144 8.49167244 - layer.0.v_cache 0.00000028 0.00022603 - layer.1.k_cache 0.00330389 0.82105146 - layer.1.v_cache 0.00000100 0.00083324 - layer.2.k_cache 0.00113731 0.41703613 - layer.2.v_cache 0.00000122 0.00121403 - layer.3.k_cache 0.00133429 0.47447244 - layer.3.v_cache 0.00000236 0.00197559 - layer.4.k_cache 0.00329518 0.87213726 - layer.4.v_cache 0.00000318 0.00325070 - layer.4.output 0.00020743 0.08874582 - ------------------------------------------------------------------------------------- - TOTAL 0.00260785 0.81706090 - (elements=1,404,928) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1404928 -Total Bytes 294148 -BPFP 1.6749 bits/point -EBPFP 3.3499 equivalent bits/point -MSE 0.817061 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 98, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8171 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample146-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample146-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample147-layer4-item1.zst (14/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample147-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 98, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 98, 128) -Output shape: (1, 98, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) -> torch.Size([1, 1, 98, 1024]) - layer.4.output: torch.Size([1, 98, 4096]) -> torch.Size([1, 1, 98, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,600B, BPFP=0.7653 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 24,368B, BPFP=1.9426 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,020B, BPFP=1.3568 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 26,304B, BPFP=2.0969 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,280B, BPFP=1.5370 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,712B, BPFP=2.1295 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,956B, BPFP=1.5909 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,492B, BPFP=2.1119 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,308B, BPFP=1.3798 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,916B, BPFP=2.1457 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 75,628B, BPFP=1.5073 -⌛️ [2/4] FRONTEND: Frontend time: 0.208s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.output: torch.Size([1, 98, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.297s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 98, 128]) - layer.0.v_cache: torch.Size([1, 8, 98, 128]) - layer.1.k_cache: torch.Size([1, 8, 98, 128]) - layer.1.v_cache: torch.Size([1, 8, 98, 128]) - layer.2.k_cache: torch.Size([1, 8, 98, 128]) - layer.2.v_cache: torch.Size([1, 8, 98, 128]) - layer.3.k_cache: torch.Size([1, 8, 98, 128]) - layer.3.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.k_cache: torch.Size([1, 8, 98, 128]) - layer.4.v_cache: torch.Size([1, 8, 98, 128]) - layer.4.output: torch.Size([1, 98, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02655566 8.63754023 - layer.0.v_cache 0.00000028 0.00023958 - layer.1.k_cache 0.00343622 0.86422239 - layer.1.v_cache 0.00000088 0.00084470 - layer.2.k_cache 0.00122497 0.41939074 - layer.2.v_cache 0.00000114 0.00119766 - layer.3.k_cache 0.00135168 0.48066237 - layer.3.v_cache 0.00000224 0.00196587 - layer.4.k_cache 0.00342672 0.89767269 - layer.4.v_cache 0.00000300 0.00316375 - layer.4.output 0.00022158 0.08438417 - ------------------------------------------------------------------------------------- - TOTAL 0.00263494 0.83174548 - (elements=1,404,928) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1404928 -Total Bytes 289584 -BPFP 1.6490 bits/point -EBPFP 3.2979 equivalent bits/point -MSE 0.831745 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.007s, Pack+Encode: 0.208s, Decode+Unpack: 0.297s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 98, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 98, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8317 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample147-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample147-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample150-layer4-item1.zst (15/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample150-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 100, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 100, 128) -Output shape: (1, 100, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) -> torch.Size([1, 1, 100, 1024]) - layer.4.output: torch.Size([1, 100, 4096]) -> torch.Size([1, 1, 100, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,800B, BPFP=0.7656 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 25,100B, BPFP=1.9609 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,964B, BPFP=1.4034 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 27,304B, BPFP=2.1331 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 20,220B, BPFP=1.5797 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 27,588B, BPFP=2.1553 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 21,004B, BPFP=1.6409 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 27,224B, BPFP=2.1269 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 18,184B, BPFP=1.4206 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 27,888B, BPFP=2.1787 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 78,700B, BPFP=1.5371 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.output: torch.Size([1, 100, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.297s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 100, 128]) - layer.0.v_cache: torch.Size([1, 8, 100, 128]) - layer.1.k_cache: torch.Size([1, 8, 100, 128]) - layer.1.v_cache: torch.Size([1, 8, 100, 128]) - layer.2.k_cache: torch.Size([1, 8, 100, 128]) - layer.2.v_cache: torch.Size([1, 8, 100, 128]) - layer.3.k_cache: torch.Size([1, 8, 100, 128]) - layer.3.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.k_cache: torch.Size([1, 8, 100, 128]) - layer.4.v_cache: torch.Size([1, 8, 100, 128]) - layer.4.output: torch.Size([1, 100, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02664840 9.58952759 - layer.0.v_cache 0.00000027 0.00022144 - layer.1.k_cache 0.00335783 0.84188629 - layer.1.v_cache 0.00000086 0.00078810 - layer.2.k_cache 0.00109031 0.40714516 - layer.2.v_cache 0.00000115 0.00113298 - layer.3.k_cache 0.00136642 0.47422359 - layer.3.v_cache 0.00000222 0.00184571 - layer.4.k_cache 0.00341075 0.89224075 - layer.4.v_cache 0.00000317 0.00308948 - layer.4.output 0.00031254 0.09144883 - ------------------------------------------------------------------------------------- - TOTAL 0.00265226 0.89842117 - (elements=1,433,600) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1433600 -Total Bytes 300976 -BPFP 1.6796 bits/point -EBPFP 3.3591 equivalent bits/point -MSE 0.898421 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.006s, Pack+Encode: 0.210s, Decode+Unpack: 0.297s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 100, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 100, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8984 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample150-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample150-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample153-layer4-item1.zst (16/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample153-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 97, 128) -Output shape: (1, 97, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.output: torch.Size([1, 97, 4096]) -> torch.Size([1, 1, 97, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,364B, BPFP=0.7542 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,680B, BPFP=1.9072 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,708B, BPFP=1.3457 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,664B, BPFP=2.0670 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,012B, BPFP=1.5312 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,088B, BPFP=2.1012 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,648B, BPFP=1.5825 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,688B, BPFP=2.0689 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,128B, BPFP=1.3795 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,556B, BPFP=2.1389 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,984B, BPFP=1.4897 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02654818 8.39822923 - layer.0.v_cache 0.00000027 0.00024450 - layer.1.k_cache 0.00347710 0.83685090 - layer.1.v_cache 0.00000089 0.00091913 - layer.2.k_cache 0.00114081 0.43420351 - layer.2.v_cache 0.00000111 0.00120679 - layer.3.k_cache 0.00140715 0.49268019 - layer.3.v_cache 0.00000211 0.00197175 - layer.4.k_cache 0.00324570 0.90120957 - layer.4.v_cache 0.00000329 0.00342746 - layer.4.output 0.00023085 0.09221516 - ------------------------------------------------------------------------------------- - TOTAL 0.00262500 0.81712883 - (elements=1,390,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1390592 -Total Bytes 283520 -BPFP 1.6311 bits/point -EBPFP 3.2622 equivalent bits/point -MSE 0.817129 ----------------------- -------------------------------------------------------- -Time: 0.511s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8171 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample153-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample153-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample154-layer4-item1.zst (17/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample154-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 89, 128) -Output shape: (1, 89, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.output: torch.Size([1, 89, 4096]) -> torch.Size([1, 1, 89, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,888B, BPFP=0.7802 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,976B, BPFP=1.9291 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,056B, BPFP=1.4094 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,184B, BPFP=2.1229 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,136B, BPFP=1.5920 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,996B, BPFP=2.1942 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 18,956B, BPFP=1.6640 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,632B, BPFP=2.1622 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,172B, BPFP=1.4196 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,088B, BPFP=2.2022 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 69,208B, BPFP=1.5188 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02697376 9.09072842 - layer.0.v_cache 0.00000027 0.00022082 - layer.1.k_cache 0.00348241 0.83463270 - layer.1.v_cache 0.00000086 0.00081447 - layer.2.k_cache 0.00112258 0.40933866 - layer.2.v_cache 0.00000113 0.00111797 - layer.3.k_cache 0.00136851 0.47865488 - layer.3.v_cache 0.00000223 0.00182848 - layer.4.k_cache 0.00328884 0.88348234 - layer.4.v_cache 0.00000300 0.00297716 - layer.4.output 0.00019035 0.08086427 - ------------------------------------------------------------------------------------- - TOTAL 0.00264321 0.85908950 - (elements=1,275,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1275904 -Total Bytes 268292 -BPFP 1.6822 bits/point -EBPFP 3.3644 equivalent bits/point -MSE 0.859090 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8591 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample154-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample154-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample157-layer4-item1.zst (18/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample157-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 104, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 104, 128) -Output shape: (1, 104, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.0.v_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.1.k_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.1.v_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.2.k_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.2.v_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.3.k_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.3.v_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.4.k_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.4.v_cache: torch.Size([1, 8, 104, 128]) -> torch.Size([1, 1, 104, 1024]) - layer.4.output: torch.Size([1, 104, 4096]) -> torch.Size([1, 1, 104, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,288B, BPFP=0.7728 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 26,620B, BPFP=1.9997 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 18,964B, BPFP=1.4246 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 28,992B, BPFP=2.1779 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,144B, BPFP=1.5883 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 29,460B, BPFP=2.2130 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 21,948B, BPFP=1.6487 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 28,868B, BPFP=2.1686 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 19,052B, BPFP=1.4312 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 29,520B, BPFP=2.2175 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 83,096B, BPFP=1.5605 -⌛️ [2/4] FRONTEND: Frontend time: 0.211s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 104, 128]) - layer.0.v_cache: torch.Size([1, 8, 104, 128]) - layer.1.k_cache: torch.Size([1, 8, 104, 128]) - layer.1.v_cache: torch.Size([1, 8, 104, 128]) - layer.2.k_cache: torch.Size([1, 8, 104, 128]) - layer.2.v_cache: torch.Size([1, 8, 104, 128]) - layer.3.k_cache: torch.Size([1, 8, 104, 128]) - layer.3.v_cache: torch.Size([1, 8, 104, 128]) - layer.4.k_cache: torch.Size([1, 8, 104, 128]) - layer.4.v_cache: torch.Size([1, 8, 104, 128]) - layer.4.output: torch.Size([1, 104, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.287s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 104, 128]) - layer.0.v_cache: torch.Size([1, 8, 104, 128]) - layer.1.k_cache: torch.Size([1, 8, 104, 128]) - layer.1.v_cache: torch.Size([1, 8, 104, 128]) - layer.2.k_cache: torch.Size([1, 8, 104, 128]) - layer.2.v_cache: torch.Size([1, 8, 104, 128]) - layer.3.k_cache: torch.Size([1, 8, 104, 128]) - layer.3.v_cache: torch.Size([1, 8, 104, 128]) - layer.4.k_cache: torch.Size([1, 8, 104, 128]) - layer.4.v_cache: torch.Size([1, 8, 104, 128]) - layer.4.output: torch.Size([1, 104, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02512770 9.54180850 - layer.0.v_cache 0.00000027 0.00023845 - layer.1.k_cache 0.00323274 0.80249192 - layer.1.v_cache 0.00000093 0.00086424 - layer.2.k_cache 0.00116098 0.40494842 - layer.2.v_cache 0.00000116 0.00117956 - layer.3.k_cache 0.00137668 0.45754191 - layer.3.v_cache 0.00000213 0.00192622 - layer.4.k_cache 0.00339989 0.90489321 - layer.4.v_cache 0.00000305 0.00315226 - layer.4.output 0.00020049 0.07625841 - ------------------------------------------------------------------------------------- - TOTAL 0.00250768 0.88743417 - (elements=1,490,944) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1490944 -Total Bytes 317952 -BPFP 1.7060 bits/point -EBPFP 3.4121 equivalent bits/point -MSE 0.887434 ----------------------- -------------------------------------------------------- -Time: 0.504s Load: 0.007s, Pack+Encode: 0.211s, Decode+Unpack: 0.287s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 104, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 104, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8874 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample157-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample157-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample159-layer4-item1.zst (19/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample159-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 82, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 82, 128) -Output shape: (1, 82, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.output: torch.Size([1, 82, 4096]) -> torch.Size([1, 1, 82, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,056B, BPFP=0.7675 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,596B, BPFP=1.9623 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,644B, BPFP=1.3952 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,444B, BPFP=2.1383 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 16,796B, BPFP=1.6002 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,244B, BPFP=2.2146 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,464B, BPFP=1.6639 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 22,852B, BPFP=2.1772 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,808B, BPFP=1.4108 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 23,468B, BPFP=2.2359 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 64,180B, BPFP=1.5287 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.output: torch.Size([1, 82, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.292s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.output: torch.Size([1, 82, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02805163 9.96949433 - layer.0.v_cache 0.00000027 0.00022646 - layer.1.k_cache 0.00355746 0.94953639 - layer.1.v_cache 0.00000072 0.00075452 - layer.2.k_cache 0.00112546 0.42756690 - layer.2.v_cache 0.00000104 0.00105651 - layer.3.k_cache 0.00143209 0.48998902 - layer.3.v_cache 0.00000205 0.00173328 - layer.4.k_cache 0.00323843 0.96616289 - layer.4.v_cache 0.00000284 0.00288109 - layer.4.output 0.00022045 0.09578477 - ------------------------------------------------------------------------------------- - TOTAL 0.00273527 0.94232432 - (elements=1,175,552) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1175552 -Total Bytes 248552 -BPFP 1.6915 bits/point -EBPFP 3.3829 equivalent bits/point -MSE 0.942324 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.006s, Pack+Encode: 0.203s, Decode+Unpack: 0.292s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 82, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9423 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample159-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample159-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample165-layer4-item1.zst (20/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample165-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 109, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 109, 128) -Output shape: (1, 109, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) -> torch.Size([1, 1, 109, 1024]) - layer.4.output: torch.Size([1, 109, 4096]) -> torch.Size([1, 1, 109, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,096B, BPFP=0.7236 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 27,536B, BPFP=1.9736 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,504B, BPFP=1.3979 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 29,912B, BPFP=2.1439 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,988B, BPFP=1.5760 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,176B, BPFP=2.1628 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,716B, BPFP=1.6282 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 29,712B, BPFP=2.1296 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 19,760B, BPFP=1.4163 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,460B, BPFP=2.1832 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 85,148B, BPFP=1.5257 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.output: torch.Size([1, 109, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.290s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 109, 128]) - layer.0.v_cache: torch.Size([1, 8, 109, 128]) - layer.1.k_cache: torch.Size([1, 8, 109, 128]) - layer.1.v_cache: torch.Size([1, 8, 109, 128]) - layer.2.k_cache: torch.Size([1, 8, 109, 128]) - layer.2.v_cache: torch.Size([1, 8, 109, 128]) - layer.3.k_cache: torch.Size([1, 8, 109, 128]) - layer.3.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.k_cache: torch.Size([1, 8, 109, 128]) - layer.4.v_cache: torch.Size([1, 8, 109, 128]) - layer.4.output: torch.Size([1, 109, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02442908 8.62844709 - layer.0.v_cache 0.00000028 0.00023056 - layer.1.k_cache 0.00329736 0.77831149 - layer.1.v_cache 0.00000085 0.00080069 - layer.2.k_cache 0.00114487 0.40848471 - layer.2.v_cache 0.00000123 0.00116568 - layer.3.k_cache 0.00134342 0.46693004 - layer.3.v_cache 0.00000206 0.00182698 - layer.4.k_cache 0.00348010 0.84797507 - layer.4.v_cache 0.00000308 0.00312938 - layer.4.output 0.00017639 0.08272059 - ------------------------------------------------------------------------------------- - TOTAL 0.00245771 0.81915600 - (elements=1,562,624) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1562624 -Total Bytes 327008 -BPFP 1.6741 bits/point -EBPFP 3.3483 equivalent bits/point -MSE 0.819156 ----------------------- -------------------------------------------------------- -Time: 0.503s Load: 0.008s, Pack+Encode: 0.205s, Decode+Unpack: 0.290s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 109, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 109, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8192 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample165-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample165-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample167-layer4-item1.zst (21/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample167-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 95, 128) -Output shape: (1, 95, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.output: torch.Size([1, 95, 4096]) -> torch.Size([1, 1, 95, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,132B, BPFP=0.7510 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,548B, BPFP=1.9365 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,068B, BPFP=1.4036 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,452B, BPFP=2.0931 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,140B, BPFP=1.5740 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,252B, BPFP=2.1589 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,956B, BPFP=1.6411 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,920B, BPFP=2.1316 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,344B, BPFP=1.4263 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,548B, BPFP=2.1832 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 75,060B, BPFP=1.5432 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02733303 9.56718365 - layer.0.v_cache 0.00000028 0.00025070 - layer.1.k_cache 0.00367247 0.83723329 - layer.1.v_cache 0.00000082 0.00088643 - layer.2.k_cache 0.00112740 0.41049700 - layer.2.v_cache 0.00000120 0.00127278 - layer.3.k_cache 0.00137809 0.47555650 - layer.3.v_cache 0.00000227 0.00199395 - layer.4.k_cache 0.00319271 0.85013355 - layer.4.v_cache 0.00000332 0.00338240 - layer.4.output 0.00022674 0.08692100 - ------------------------------------------------------------------------------------- - TOTAL 0.00268704 0.89257673 - (elements=1,361,920) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1361920 -Total Bytes 285420 -BPFP 1.6766 bits/point -EBPFP 3.3531 equivalent bits/point -MSE 0.892577 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.006s, Pack+Encode: 0.204s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8926 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample167-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample167-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample168-layer4-item1.zst (22/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample168-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 103, 128) -Output shape: (1, 103, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.output: torch.Size([1, 103, 4096]) -> torch.Size([1, 1, 103, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,196B, BPFP=0.7734 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 26,380B, BPFP=2.0009 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 18,640B, BPFP=1.4138 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 28,476B, BPFP=2.1599 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,012B, BPFP=1.5938 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 28,788B, BPFP=2.1836 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 21,600B, BPFP=1.6383 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 28,288B, BPFP=2.1456 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 18,784B, BPFP=1.4248 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 29,044B, BPFP=2.2030 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 81,080B, BPFP=1.5375 -⌛️ [2/4] FRONTEND: Frontend time: 0.206s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.287s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.03002706 9.01381291 - layer.0.v_cache 0.00000028 0.00022587 - layer.1.k_cache 0.00329593 0.79242551 - layer.1.v_cache 0.00000087 0.00081840 - layer.2.k_cache 0.00114743 0.39720613 - layer.2.v_cache 0.00000126 0.00111170 - layer.3.k_cache 0.00133553 0.44942330 - layer.3.v_cache 0.00000217 0.00187167 - layer.4.k_cache 0.00338515 0.84470175 - layer.4.v_cache 0.00000318 0.00317392 - layer.4.output 0.00017408 0.07585537 - ------------------------------------------------------------------------------------- - TOTAL 0.00284966 0.84344233 - (elements=1,476,608) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1476608 -Total Bytes 312288 -BPFP 1.6919 bits/point -EBPFP 3.3838 equivalent bits/point -MSE 0.843442 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.007s, Pack+Encode: 0.206s, Decode+Unpack: 0.287s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8434 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample168-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample168-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample174-layer4-item1.zst (23/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample174-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 103, 128) -Output shape: (1, 103, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.output: torch.Size([1, 103, 4096]) -> torch.Size([1, 1, 103, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,296B, BPFP=0.7809 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 26,372B, BPFP=2.0003 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 18,652B, BPFP=1.4147 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 28,416B, BPFP=2.1553 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,124B, BPFP=1.6022 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 28,948B, BPFP=2.1957 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 21,800B, BPFP=1.6535 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 28,524B, BPFP=2.1635 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 18,804B, BPFP=1.4263 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 29,236B, BPFP=2.2175 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 82,468B, BPFP=1.5638 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.300s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02636531 9.57137558 - layer.0.v_cache 0.00000028 0.00022810 - layer.1.k_cache 0.00340973 0.84353356 - layer.1.v_cache 0.00000088 0.00080123 - layer.2.k_cache 0.00113950 0.40540291 - layer.2.v_cache 0.00000120 0.00114791 - layer.3.k_cache 0.00133761 0.46236997 - layer.3.v_cache 0.00000227 0.00185175 - layer.4.k_cache 0.00339299 0.86312533 - layer.4.v_cache 0.00000316 0.00313638 - layer.4.output 0.00023648 0.07913694 - ------------------------------------------------------------------------------------- - TOTAL 0.00261420 0.89068003 - (elements=1,476,608) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1476608 -Total Bytes 314640 -BPFP 1.7047 bits/point -EBPFP 3.4093 equivalent bits/point -MSE 0.890680 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.007s, Pack+Encode: 0.205s, Decode+Unpack: 0.300s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8907 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample174-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample174-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample182-layer4-item1.zst (24/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample182-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 92, 128) -Output shape: (1, 92, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.output: torch.Size([1, 92, 4096]) -> torch.Size([1, 1, 92, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,192B, BPFP=0.7806 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,336B, BPFP=1.9817 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,892B, BPFP=1.4344 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,468B, BPFP=2.1627 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,056B, BPFP=1.6182 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,888B, BPFP=2.1984 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,948B, BPFP=1.6940 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,780B, BPFP=2.1892 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,144B, BPFP=1.4558 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,240B, BPFP=2.2283 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,856B, BPFP=1.5679 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02458628 8.96046779 - layer.0.v_cache 0.00000028 0.00023505 - layer.1.k_cache 0.00347829 0.86617976 - layer.1.v_cache 0.00000086 0.00081966 - layer.2.k_cache 0.00113727 0.39630828 - layer.2.v_cache 0.00000116 0.00112744 - layer.3.k_cache 0.00132770 0.46726799 - layer.3.v_cache 0.00000243 0.00185272 - layer.4.k_cache 0.00339865 0.85811026 - layer.4.v_cache 0.00000297 0.00297973 - layer.4.output 0.00018871 0.08642903 - ------------------------------------------------------------------------------------- - TOTAL 0.00247791 0.85007606 - (elements=1,318,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1318912 -Total Bytes 282800 -BPFP 1.7154 bits/point -EBPFP 3.4307 equivalent bits/point -MSE 0.850076 ----------------------- -------------------------------------------------------- -Time: 0.515s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8501 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample182-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample182-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample185-layer4-item1.zst (25/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample185-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 96, 128) -Output shape: (1, 96, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.output: torch.Size([1, 96, 4096]) -> torch.Size([1, 1, 96, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,372B, BPFP=0.7627 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,540B, BPFP=1.9157 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,836B, BPFP=1.3701 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,400B, BPFP=2.0671 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,900B, BPFP=1.5381 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,916B, BPFP=2.1090 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,868B, BPFP=1.6169 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,996B, BPFP=2.1156 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,084B, BPFP=1.3903 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,340B, BPFP=2.1436 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 76,144B, BPFP=1.5492 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02634023 8.63892110 - layer.0.v_cache 0.00000027 0.00022330 - layer.1.k_cache 0.00351230 0.83090146 - layer.1.v_cache 0.00000075 0.00079098 - layer.2.k_cache 0.00114531 0.39917850 - layer.2.v_cache 0.00000110 0.00110211 - layer.3.k_cache 0.00136477 0.46863747 - layer.3.v_cache 0.00000217 0.00188588 - layer.4.k_cache 0.00356451 0.86840439 - layer.4.v_cache 0.00000291 0.00302172 - layer.4.output 0.00018809 0.08683455 - ------------------------------------------------------------------------------------- - TOTAL 0.00262048 0.82574322 - (elements=1,376,256) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1376256 -Total Bytes 285396 -BPFP 1.6590 bits/point -EBPFP 3.3179 equivalent bits/point -MSE 0.825743 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.006s, Pack+Encode: 0.210s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8257 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample185-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample185-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample191-layer4-item1.zst (26/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample191-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 93, 128) -Output shape: (1, 93, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.output: torch.Size([1, 93, 4096]) -> torch.Size([1, 1, 93, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,268B, BPFP=0.7786 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,488B, BPFP=1.9731 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,804B, BPFP=1.4116 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,404B, BPFP=2.1341 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,316B, BPFP=1.6226 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,988B, BPFP=2.1831 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,772B, BPFP=1.6610 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,636B, BPFP=2.1536 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,216B, BPFP=1.4462 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,464B, BPFP=2.2231 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 74,236B, BPFP=1.5591 -⌛️ [2/4] FRONTEND: Frontend time: 0.217s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.308s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02618607 8.81040512 - layer.0.v_cache 0.00000028 0.00023089 - layer.1.k_cache 0.00358459 0.85735854 - layer.1.v_cache 0.00000078 0.00080429 - layer.2.k_cache 0.00113483 0.41530297 - layer.2.v_cache 0.00000114 0.00115141 - layer.3.k_cache 0.00136506 0.48415555 - layer.3.v_cache 0.00000214 0.00188974 - layer.4.k_cache 0.00334622 0.87478490 - layer.4.v_cache 0.00000303 0.00307359 - layer.4.output 0.00022383 0.09224577 - ------------------------------------------------------------------------------------- - TOTAL 0.00260854 0.84415286 - (elements=1,333,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1333248 -Total Bytes 283592 -BPFP 1.7017 bits/point -EBPFP 3.4033 equivalent bits/point -MSE 0.844153 ----------------------- -------------------------------------------------------- -Time: 0.533s Load: 0.007s, Pack+Encode: 0.217s, Decode+Unpack: 0.308s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8442 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample191-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample191-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample196-layer4-item1.zst (27/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample196-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 96, 128) -Output shape: (1, 96, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.output: torch.Size([1, 96, 4096]) -> torch.Size([1, 1, 96, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,436B, BPFP=0.7679 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,772B, BPFP=1.9346 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,932B, BPFP=1.3779 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,784B, BPFP=2.0983 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,300B, BPFP=1.5706 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,328B, BPFP=2.1426 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,760B, BPFP=1.6081 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,916B, BPFP=2.1090 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,252B, BPFP=1.4040 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,440B, BPFP=2.1517 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 75,588B, BPFP=1.5378 -⌛️ [2/4] FRONTEND: Frontend time: 0.223s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.307s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02641018 8.34279251 - layer.0.v_cache 0.00000028 0.00023202 - layer.1.k_cache 0.00334712 0.82118678 - layer.1.v_cache 0.00000079 0.00085753 - layer.2.k_cache 0.00118939 0.41674765 - layer.2.v_cache 0.00000116 0.00124040 - layer.3.k_cache 0.00138481 0.46456162 - layer.3.v_cache 0.00000224 0.00194597 - layer.4.k_cache 0.00336247 0.82522567 - layer.4.v_cache 0.00000319 0.00330839 - layer.4.output 0.00019453 0.07299321 - ------------------------------------------------------------------------------------- - TOTAL 0.00260570 0.79786224 - (elements=1,376,256) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1376256 -Total Bytes 286508 -BPFP 1.6654 bits/point -EBPFP 3.3309 equivalent bits/point -MSE 0.797862 ----------------------- -------------------------------------------------------- -Time: 0.537s Load: 0.007s, Pack+Encode: 0.223s, Decode+Unpack: 0.307s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7979 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample196-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample196-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample197-layer4-item1.zst (28/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample197-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 92, 128) -Output shape: (1, 92, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.output: torch.Size([1, 92, 4096]) -> torch.Size([1, 1, 92, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,236B, BPFP=0.7843 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,016B, BPFP=1.9545 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,860B, BPFP=1.4317 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,364B, BPFP=2.1539 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,292B, BPFP=1.6382 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,052B, BPFP=2.2123 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,864B, BPFP=1.6868 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,516B, BPFP=2.1668 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,196B, BPFP=1.4603 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,104B, BPFP=2.2167 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,148B, BPFP=1.5529 -⌛️ [2/4] FRONTEND: Frontend time: 0.216s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02542763 9.31790559 - layer.0.v_cache 0.00000028 0.00023446 - layer.1.k_cache 0.00330708 0.83873566 - layer.1.v_cache 0.00000075 0.00082690 - layer.2.k_cache 0.00113337 0.41600144 - layer.2.v_cache 0.00000121 0.00119881 - layer.3.k_cache 0.00137180 0.47507647 - layer.3.v_cache 0.00000212 0.00192658 - layer.4.k_cache 0.00335567 0.86708807 - layer.4.v_cache 0.00000315 0.00323999 - layer.4.output 0.00020664 0.08036359 - ------------------------------------------------------------------------------------- - TOTAL 0.00253069 0.87454917 - (elements=1,318,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1318912 -Total Bytes 281648 -BPFP 1.7084 bits/point -EBPFP 3.4167 equivalent bits/point -MSE 0.874549 ----------------------- -------------------------------------------------------- -Time: 0.520s Load: 0.006s, Pack+Encode: 0.216s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8745 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample197-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample197-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample2-layer4-item1.zst (29/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample2-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.010s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 178, 128) -Output shape: (1, 178, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.0.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.1.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.1.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.2.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.2.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.3.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.3.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.4.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.4.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) - layer.4.output: torch.Size([1, 178, 4096]) -> torch.Size([1, 1, 178, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 16,540B, BPFP=0.7259 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 43,100B, BPFP=1.8917 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 30,148B, BPFP=1.3232 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 45,792B, BPFP=2.0098 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 33,960B, BPFP=1.4905 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 46,516B, BPFP=2.0416 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 35,280B, BPFP=1.5485 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 46,112B, BPFP=2.0239 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 30,516B, BPFP=1.3394 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 46,924B, BPFP=2.0595 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 132,152B, BPFP=1.4501 -⌛️ [2/4] FRONTEND: Frontend time: 0.264s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 178, 128]) - layer.0.v_cache: torch.Size([1, 8, 178, 128]) - layer.1.k_cache: torch.Size([1, 8, 178, 128]) - layer.1.v_cache: torch.Size([1, 8, 178, 128]) - layer.2.k_cache: torch.Size([1, 8, 178, 128]) - layer.2.v_cache: torch.Size([1, 8, 178, 128]) - layer.3.k_cache: torch.Size([1, 8, 178, 128]) - layer.3.v_cache: torch.Size([1, 8, 178, 128]) - layer.4.k_cache: torch.Size([1, 8, 178, 128]) - layer.4.v_cache: torch.Size([1, 8, 178, 128]) - layer.4.output: torch.Size([1, 178, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.405s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 178, 128]) - layer.0.v_cache: torch.Size([1, 8, 178, 128]) - layer.1.k_cache: torch.Size([1, 8, 178, 128]) - layer.1.v_cache: torch.Size([1, 8, 178, 128]) - layer.2.k_cache: torch.Size([1, 8, 178, 128]) - layer.2.v_cache: torch.Size([1, 8, 178, 128]) - layer.3.k_cache: torch.Size([1, 8, 178, 128]) - layer.3.v_cache: torch.Size([1, 8, 178, 128]) - layer.4.k_cache: torch.Size([1, 8, 178, 128]) - layer.4.v_cache: torch.Size([1, 8, 178, 128]) - layer.4.output: torch.Size([1, 178, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02409235 7.42891736 - layer.0.v_cache 0.00000026 0.00021943 - layer.1.k_cache 0.00309449 0.69218020 - layer.1.v_cache 0.00000082 0.00078660 - layer.2.k_cache 0.00118451 0.39601260 - layer.2.v_cache 0.00000118 0.00110626 - layer.3.k_cache 0.00136933 0.45109927 - layer.3.v_cache 0.00000226 0.00187067 - layer.4.k_cache 0.00340187 0.80474776 - layer.4.v_cache 0.00000300 0.00295725 - layer.4.output 0.00017401 0.06430830 - ------------------------------------------------------------------------------------- - TOTAL 0.00241758 0.71693790 - (elements=2,551,808) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2551808 -Total Bytes 507040 -BPFP 1.5896 bits/point -EBPFP 3.1792 equivalent bits/point -MSE 0.716938 ----------------------- -------------------------------------------------------- -Time: 0.679s Load: 0.010s, Pack+Encode: 0.264s, Decode+Unpack: 0.405s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7169 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample2-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample2-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample201-layer4-item1.zst (30/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample201-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 95, 128) -Output shape: (1, 95, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) -> torch.Size([1, 1, 95, 1024]) - layer.4.output: torch.Size([1, 95, 4096]) -> torch.Size([1, 1, 95, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,092B, BPFP=0.7477 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,344B, BPFP=1.9197 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,180B, BPFP=1.4128 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,380B, BPFP=2.0872 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,572B, BPFP=1.6095 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,028B, BPFP=2.1405 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,072B, BPFP=1.6507 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,456B, BPFP=2.0934 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,344B, BPFP=1.4263 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,120B, BPFP=2.1480 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 74,020B, BPFP=1.5218 -⌛️ [2/4] FRONTEND: Frontend time: 0.212s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.299s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 95, 128]) - layer.0.v_cache: torch.Size([1, 8, 95, 128]) - layer.1.k_cache: torch.Size([1, 8, 95, 128]) - layer.1.v_cache: torch.Size([1, 8, 95, 128]) - layer.2.k_cache: torch.Size([1, 8, 95, 128]) - layer.2.v_cache: torch.Size([1, 8, 95, 128]) - layer.3.k_cache: torch.Size([1, 8, 95, 128]) - layer.3.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.k_cache: torch.Size([1, 8, 95, 128]) - layer.4.v_cache: torch.Size([1, 8, 95, 128]) - layer.4.output: torch.Size([1, 95, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02560628 8.79905428 - layer.0.v_cache 0.00000028 0.00022873 - layer.1.k_cache 0.00332260 0.86858842 - layer.1.v_cache 0.00000077 0.00083978 - layer.2.k_cache 0.00114542 0.42579980 - layer.2.v_cache 0.00000120 0.00116849 - layer.3.k_cache 0.00143781 0.48452201 - layer.3.v_cache 0.00000216 0.00185109 - layer.4.k_cache 0.00324414 0.90596362 - layer.4.v_cache 0.00000298 0.00298450 - layer.4.output 0.00027467 0.09089894 - ------------------------------------------------------------------------------------- - TOTAL 0.00256159 0.84675689 - (elements=1,361,920) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1361920 -Total Bytes 283608 -BPFP 1.6659 bits/point -EBPFP 3.3319 equivalent bits/point -MSE 0.846757 ----------------------- -------------------------------------------------------- -Time: 0.516s Load: 0.006s, Pack+Encode: 0.212s, Decode+Unpack: 0.299s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 95, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 95, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8468 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample201-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample201-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample212-layer4-item1.zst (31/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample212-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 94, 128) -Output shape: (1, 94, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.output: torch.Size([1, 94, 4096]) -> torch.Size([1, 1, 94, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,224B, BPFP=0.7666 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,436B, BPFP=1.9478 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,144B, BPFP=1.4249 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,276B, BPFP=2.1007 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,468B, BPFP=1.6180 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,984B, BPFP=2.1596 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,088B, BPFP=1.6695 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,844B, BPFP=2.1479 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,256B, BPFP=1.4342 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,312B, BPFP=2.1868 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,832B, BPFP=1.5341 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02535593 8.39036138 - layer.0.v_cache 0.00000029 0.00021995 - layer.1.k_cache 0.00347189 0.83858263 - layer.1.v_cache 0.00000073 0.00075098 - layer.2.k_cache 0.00115370 0.39896823 - layer.2.v_cache 0.00000109 0.00109513 - layer.3.k_cache 0.00136415 0.46520404 - layer.3.v_cache 0.00000218 0.00181022 - layer.4.k_cache 0.00350239 0.85282322 - layer.4.v_cache 0.00000301 0.00307067 - layer.4.output 0.00017689 0.09335934 - ------------------------------------------------------------------------------------- - TOTAL 0.00254021 0.80902313 - (elements=1,347,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1347584 -Total Bytes 283864 -BPFP 1.6852 bits/point -EBPFP 3.3703 equivalent bits/point -MSE 0.809023 ----------------------- -------------------------------------------------------- -Time: 0.515s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8090 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample212-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample212-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample214-layer4-item1.zst (32/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample214-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 99, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 99, 128) -Output shape: (1, 99, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.output: torch.Size([1, 99, 4096]) -> torch.Size([1, 1, 99, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,456B, BPFP=0.7462 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 24,492B, BPFP=1.9328 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,444B, BPFP=1.3766 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 26,772B, BPFP=2.1127 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,812B, BPFP=1.5634 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 27,228B, BPFP=2.1487 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,492B, BPFP=1.6171 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,764B, BPFP=2.1121 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,772B, BPFP=1.4025 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 27,452B, BPFP=2.1664 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 76,104B, BPFP=1.5014 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.output: torch.Size([1, 99, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.300s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.output: torch.Size([1, 99, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02440176 7.97288744 - layer.0.v_cache 0.00000028 0.00022947 - layer.1.k_cache 0.00326434 0.83401505 - layer.1.v_cache 0.00000081 0.00077908 - layer.2.k_cache 0.00115982 0.42109110 - layer.2.v_cache 0.00000131 0.00112911 - layer.3.k_cache 0.00139121 0.48312262 - layer.3.v_cache 0.00000205 0.00180836 - layer.4.k_cache 0.00335893 0.89481539 - layer.4.v_cache 0.00000299 0.00317244 - layer.4.output 0.00020438 0.08623992 - ------------------------------------------------------------------------------------- - TOTAL 0.00245722 0.78271498 - (elements=1,419,264) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1419264 -Total Bytes 293788 -BPFP 1.6560 bits/point -EBPFP 3.3120 equivalent bits/point -MSE 0.782715 ----------------------- -------------------------------------------------------- -Time: 0.516s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.300s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 99, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7827 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample214-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample214-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample224-layer4-item1.zst (33/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample224-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 103, 128) -Output shape: (1, 103, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) -> torch.Size([1, 1, 103, 1024]) - layer.4.output: torch.Size([1, 103, 4096]) -> torch.Size([1, 1, 103, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,052B, BPFP=0.7624 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 26,080B, BPFP=1.9782 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 18,780B, BPFP=1.4245 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 28,232B, BPFP=2.1414 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 21,028B, BPFP=1.5950 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 28,656B, BPFP=2.1735 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 21,928B, BPFP=1.6632 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 28,176B, BPFP=2.1371 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 18,860B, BPFP=1.4305 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 28,804B, BPFP=2.1848 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 81,748B, BPFP=1.5501 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.300s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 103, 128]) - layer.0.v_cache: torch.Size([1, 8, 103, 128]) - layer.1.k_cache: torch.Size([1, 8, 103, 128]) - layer.1.v_cache: torch.Size([1, 8, 103, 128]) - layer.2.k_cache: torch.Size([1, 8, 103, 128]) - layer.2.v_cache: torch.Size([1, 8, 103, 128]) - layer.3.k_cache: torch.Size([1, 8, 103, 128]) - layer.3.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.k_cache: torch.Size([1, 8, 103, 128]) - layer.4.v_cache: torch.Size([1, 8, 103, 128]) - layer.4.output: torch.Size([1, 103, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02603974 9.59222353 - layer.0.v_cache 0.00000026 0.00021785 - layer.1.k_cache 0.00325535 0.84833845 - layer.1.v_cache 0.00000093 0.00080141 - layer.2.k_cache 0.00112553 0.40892366 - layer.2.v_cache 0.00000114 0.00109245 - layer.3.k_cache 0.00141218 0.47342371 - layer.3.v_cache 0.00000204 0.00181121 - layer.4.k_cache 0.00324713 0.90190465 - layer.4.v_cache 0.00000302 0.00304244 - layer.4.output 0.00020093 0.08909076 - ------------------------------------------------------------------------------------- - TOTAL 0.00256365 0.89915303 - (elements=1,476,608) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1476608 -Total Bytes 312344 -BPFP 1.6922 bits/point -EBPFP 3.3844 equivalent bits/point -MSE 0.899153 ----------------------- -------------------------------------------------------- -Time: 0.516s Load: 0.006s, Pack+Encode: 0.209s, Decode+Unpack: 0.300s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 103, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 103, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8992 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample224-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample224-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample227-layer4-item1.zst (34/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample227-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 96, 128) -Output shape: (1, 96, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) -> torch.Size([1, 1, 96, 1024]) - layer.4.output: torch.Size([1, 96, 4096]) -> torch.Size([1, 1, 96, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,924B, BPFP=0.8076 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,844B, BPFP=1.9404 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,224B, BPFP=1.4017 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,668B, BPFP=2.0889 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,444B, BPFP=1.5824 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,312B, BPFP=2.1413 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,108B, BPFP=1.6364 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,140B, BPFP=2.1273 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,536B, BPFP=1.4271 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,680B, BPFP=2.1712 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 75,688B, BPFP=1.5399 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 96, 128]) - layer.0.v_cache: torch.Size([1, 8, 96, 128]) - layer.1.k_cache: torch.Size([1, 8, 96, 128]) - layer.1.v_cache: torch.Size([1, 8, 96, 128]) - layer.2.k_cache: torch.Size([1, 8, 96, 128]) - layer.2.v_cache: torch.Size([1, 8, 96, 128]) - layer.3.k_cache: torch.Size([1, 8, 96, 128]) - layer.3.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.k_cache: torch.Size([1, 8, 96, 128]) - layer.4.v_cache: torch.Size([1, 8, 96, 128]) - layer.4.output: torch.Size([1, 96, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02591230 8.62229411 - layer.0.v_cache 0.00000028 0.00021768 - layer.1.k_cache 0.00321835 0.80863237 - layer.1.v_cache 0.00000081 0.00086501 - layer.2.k_cache 0.00115341 0.40979377 - layer.2.v_cache 0.00000115 0.00114236 - layer.3.k_cache 0.00138578 0.46226505 - layer.3.v_cache 0.00000239 0.00196140 - layer.4.k_cache 0.00338578 0.83702739 - layer.4.v_cache 0.00000322 0.00329243 - layer.4.output 0.00018605 0.08113537 - ------------------------------------------------------------------------------------- - TOTAL 0.00255769 0.81943093 - (elements=1,376,256) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1376256 -Total Bytes 288568 -BPFP 1.6774 bits/point -EBPFP 3.3548 equivalent bits/point -MSE 0.819431 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 96, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 96, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8194 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample227-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample227-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample233-layer4-item1.zst (35/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample233-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 85, 128) -Output shape: (1, 85, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.output: torch.Size([1, 85, 4096]) -> torch.Size([1, 1, 85, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,864B, BPFP=0.8147 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,756B, BPFP=1.9996 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,328B, BPFP=1.4088 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 23,988B, BPFP=2.2048 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,564B, BPFP=1.6143 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,688B, BPFP=2.2691 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 18,164B, BPFP=1.6695 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,388B, BPFP=2.2415 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,676B, BPFP=1.4408 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 24,812B, BPFP=2.2805 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 70,564B, BPFP=1.6214 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.297s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02725663 9.68375819 - layer.0.v_cache 0.00000027 0.00024173 - layer.1.k_cache 0.00328410 0.90437281 - layer.1.v_cache 0.00000085 0.00087641 - layer.2.k_cache 0.00114159 0.44561979 - layer.2.v_cache 0.00000125 0.00128137 - layer.3.k_cache 0.00133976 0.49526022 - layer.3.v_cache 0.00000266 0.00209650 - layer.4.k_cache 0.00332377 0.92088695 - layer.4.v_cache 0.00000336 0.00348818 - layer.4.output 0.00023926 0.09477260 - ------------------------------------------------------------------------------------- - TOTAL 0.00266509 0.91692661 - (elements=1,218,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1218560 -Total Bytes 265792 -BPFP 1.7450 bits/point -EBPFP 3.4899 equivalent bits/point -MSE 0.916927 ----------------------- -------------------------------------------------------- -Time: 0.514s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.297s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9169 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample233-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample233-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample241-layer4-item1.zst (36/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample241-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 97, 128) -Output shape: (1, 97, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.output: torch.Size([1, 97, 4096]) -> torch.Size([1, 1, 97, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,060B, BPFP=0.7297 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,552B, BPFP=1.8969 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,904B, BPFP=1.3615 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,552B, BPFP=2.0580 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,036B, BPFP=1.5332 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,120B, BPFP=2.1037 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,708B, BPFP=1.5873 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,592B, BPFP=2.0612 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,972B, BPFP=1.3669 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,312B, BPFP=2.1192 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 74,216B, BPFP=1.4944 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02553191 8.34447676 - layer.0.v_cache 0.00000029 0.00024866 - layer.1.k_cache 0.00322990 0.81368326 - layer.1.v_cache 0.00000089 0.00084857 - layer.2.k_cache 0.00117012 0.41652510 - layer.2.v_cache 0.00000108 0.00118112 - layer.3.k_cache 0.00135684 0.47318897 - layer.3.v_cache 0.00000209 0.00191762 - layer.4.k_cache 0.00331928 0.85838931 - layer.4.v_cache 0.00000299 0.00314493 - layer.4.output 0.00021425 0.08700518 - ------------------------------------------------------------------------------------- - TOTAL 0.00253374 0.80440179 - (elements=1,390,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1390592 -Total Bytes 283024 -BPFP 1.6282 bits/point -EBPFP 3.2564 equivalent bits/point -MSE 0.804402 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8044 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample241-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample241-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample250-layer4-item1.zst (37/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample250-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 97, 128) -Output shape: (1, 97, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) -> torch.Size([1, 1, 97, 1024]) - layer.4.output: torch.Size([1, 97, 4096]) -> torch.Size([1, 1, 97, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,460B, BPFP=0.7619 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,764B, BPFP=1.9140 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,080B, BPFP=1.3756 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,940B, BPFP=2.0892 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,212B, BPFP=1.5474 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,440B, BPFP=2.1295 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,876B, BPFP=1.6008 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,280B, BPFP=2.1166 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,216B, BPFP=1.3866 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,804B, BPFP=2.1588 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 78,224B, BPFP=1.5751 -⌛️ [2/4] FRONTEND: Frontend time: 0.213s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 97, 128]) - layer.0.v_cache: torch.Size([1, 8, 97, 128]) - layer.1.k_cache: torch.Size([1, 8, 97, 128]) - layer.1.v_cache: torch.Size([1, 8, 97, 128]) - layer.2.k_cache: torch.Size([1, 8, 97, 128]) - layer.2.v_cache: torch.Size([1, 8, 97, 128]) - layer.3.k_cache: torch.Size([1, 8, 97, 128]) - layer.3.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.k_cache: torch.Size([1, 8, 97, 128]) - layer.4.v_cache: torch.Size([1, 8, 97, 128]) - layer.4.output: torch.Size([1, 97, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02408928 8.52254903 - layer.0.v_cache 0.00000028 0.00025437 - layer.1.k_cache 0.00330002 0.79464431 - layer.1.v_cache 0.00000090 0.00082465 - layer.2.k_cache 0.00113900 0.40602883 - layer.2.v_cache 0.00000130 0.00117890 - layer.3.k_cache 0.00136322 0.46068313 - layer.3.v_cache 0.00000251 0.00206107 - layer.4.k_cache 0.00342626 0.85458154 - layer.4.v_cache 0.00000319 0.00318730 - layer.4.output 0.00021399 0.08945499 - ------------------------------------------------------------------------------------- - TOTAL 0.00244156 0.81455808 - (elements=1,390,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1390592 -Total Bytes 290296 -BPFP 1.6701 bits/point -EBPFP 3.3401 equivalent bits/point -MSE 0.814558 ----------------------- -------------------------------------------------------- -Time: 0.515s Load: 0.006s, Pack+Encode: 0.213s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 97, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 97, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8146 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample250-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample250-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample251-layer4-item1.zst (38/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample251-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 99, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 99, 128) -Output shape: (1, 99, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) -> torch.Size([1, 1, 99, 1024]) - layer.4.output: torch.Size([1, 99, 4096]) -> torch.Size([1, 1, 99, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,172B, BPFP=0.7238 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 24,488B, BPFP=1.9324 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,420B, BPFP=1.3747 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 26,560B, BPFP=2.0960 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,624B, BPFP=1.5486 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 26,912B, BPFP=2.1237 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 20,396B, BPFP=1.6095 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 26,516B, BPFP=2.0925 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,628B, BPFP=1.3911 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 27,208B, BPFP=2.1471 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 76,736B, BPFP=1.5139 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.output: torch.Size([1, 99, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.300s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 99, 128]) - layer.0.v_cache: torch.Size([1, 8, 99, 128]) - layer.1.k_cache: torch.Size([1, 8, 99, 128]) - layer.1.v_cache: torch.Size([1, 8, 99, 128]) - layer.2.k_cache: torch.Size([1, 8, 99, 128]) - layer.2.v_cache: torch.Size([1, 8, 99, 128]) - layer.3.k_cache: torch.Size([1, 8, 99, 128]) - layer.3.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.k_cache: torch.Size([1, 8, 99, 128]) - layer.4.v_cache: torch.Size([1, 8, 99, 128]) - layer.4.output: torch.Size([1, 99, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02490362 7.99439093 - layer.0.v_cache 0.00000029 0.00023659 - layer.1.k_cache 0.00335574 0.87376435 - layer.1.v_cache 0.00000081 0.00076597 - layer.2.k_cache 0.00112037 0.40546579 - layer.2.v_cache 0.00000110 0.00104381 - layer.3.k_cache 0.00138480 0.46416473 - layer.3.v_cache 0.00000194 0.00167598 - layer.4.k_cache 0.00346640 0.87850251 - layer.4.v_cache 0.00000296 0.00284821 - layer.4.output 0.00019387 0.08249639 - ------------------------------------------------------------------------------------- - TOTAL 0.00250096 0.78234603 - (elements=1,419,264) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1419264 -Total Bytes 292660 -BPFP 1.6496 bits/point -EBPFP 3.2993 equivalent bits/point -MSE 0.782346 ----------------------- -------------------------------------------------------- -Time: 0.516s Load: 0.007s, Pack+Encode: 0.210s, Decode+Unpack: 0.300s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 99, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 99, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7823 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample251-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample251-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample257-layer4-item1.zst (39/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample257-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 86, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 86, 128) -Output shape: (1, 86, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.0.v_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.1.k_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.1.v_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.2.k_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.2.v_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.3.k_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.3.v_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.4.k_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.4.v_cache: torch.Size([1, 8, 86, 128]) -> torch.Size([1, 1, 86, 1024]) - layer.4.output: torch.Size([1, 86, 4096]) -> torch.Size([1, 1, 86, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,140B, BPFP=0.8303 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,984B, BPFP=1.9971 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,680B, BPFP=1.4244 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,300B, BPFP=2.2075 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,088B, BPFP=1.6432 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,352B, BPFP=2.3031 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 18,620B, BPFP=1.6915 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,968B, BPFP=2.2682 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,712B, BPFP=1.4273 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,128B, BPFP=2.2827 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 70,656B, BPFP=1.6047 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 86, 128]) - layer.0.v_cache: torch.Size([1, 8, 86, 128]) - layer.1.k_cache: torch.Size([1, 8, 86, 128]) - layer.1.v_cache: torch.Size([1, 8, 86, 128]) - layer.2.k_cache: torch.Size([1, 8, 86, 128]) - layer.2.v_cache: torch.Size([1, 8, 86, 128]) - layer.3.k_cache: torch.Size([1, 8, 86, 128]) - layer.3.v_cache: torch.Size([1, 8, 86, 128]) - layer.4.k_cache: torch.Size([1, 8, 86, 128]) - layer.4.v_cache: torch.Size([1, 8, 86, 128]) - layer.4.output: torch.Size([1, 86, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 86, 128]) - layer.0.v_cache: torch.Size([1, 8, 86, 128]) - layer.1.k_cache: torch.Size([1, 8, 86, 128]) - layer.1.v_cache: torch.Size([1, 8, 86, 128]) - layer.2.k_cache: torch.Size([1, 8, 86, 128]) - layer.2.v_cache: torch.Size([1, 8, 86, 128]) - layer.3.k_cache: torch.Size([1, 8, 86, 128]) - layer.3.v_cache: torch.Size([1, 8, 86, 128]) - layer.4.k_cache: torch.Size([1, 8, 86, 128]) - layer.4.v_cache: torch.Size([1, 8, 86, 128]) - layer.4.output: torch.Size([1, 86, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02609516 8.93383860 - layer.0.v_cache 0.00000027 0.00022490 - layer.1.k_cache 0.00348465 0.88446781 - layer.1.v_cache 0.00000080 0.00084105 - layer.2.k_cache 0.00118276 0.42168045 - layer.2.v_cache 0.00000133 0.00130245 - layer.3.k_cache 0.00137135 0.50197100 - layer.3.v_cache 0.00000258 0.00207655 - layer.4.k_cache 0.00338397 0.89090782 - layer.4.v_cache 0.00000307 0.00328391 - layer.4.output 0.00019889 0.09062282 - ------------------------------------------------------------------------------------- - TOTAL 0.00259439 0.85736327 - (elements=1,232,896) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1232896 -Total Bytes 269628 -BPFP 1.7496 bits/point -EBPFP 3.4991 equivalent bits/point -MSE 0.857363 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.006s, Pack+Encode: 0.210s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 86, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 86, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8574 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample257-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample257-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample258-layer4-item1.zst (40/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample258-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 89, 128) -Output shape: (1, 89, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.output: torch.Size([1, 89, 4096]) -> torch.Size([1, 1, 89, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,236B, BPFP=0.8107 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 22,256B, BPFP=1.9537 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,008B, BPFP=1.4052 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,116B, BPFP=2.1169 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,288B, BPFP=1.6053 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,992B, BPFP=2.1938 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,008B, BPFP=1.6685 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,788B, BPFP=2.1759 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,212B, BPFP=1.4231 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,304B, BPFP=2.2212 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 71,800B, BPFP=1.5757 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02681794 8.90518154 - layer.0.v_cache 0.00000027 0.00024013 - layer.1.k_cache 0.00344681 0.85114691 - layer.1.v_cache 0.00000079 0.00087214 - layer.2.k_cache 0.00114729 0.44201750 - layer.2.v_cache 0.00000121 0.00123149 - layer.3.k_cache 0.00138721 0.50507295 - layer.3.v_cache 0.00000216 0.00204487 - layer.4.k_cache 0.00341441 0.89050122 - layer.4.v_cache 0.00000316 0.00339639 - layer.4.output 0.00022172 0.09198880 - ------------------------------------------------------------------------------------- - TOTAL 0.00265058 0.85497574 - (elements=1,275,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1275904 -Total Bytes 272008 -BPFP 1.7055 bits/point -EBPFP 3.4110 equivalent bits/point -MSE 0.854976 ----------------------- -------------------------------------------------------- -Time: 0.510s Load: 0.006s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8550 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample258-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample258-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample263-layer4-item1.zst (41/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample263-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 89, 128) -Output shape: (1, 89, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) -> torch.Size([1, 1, 89, 1024]) - layer.4.output: torch.Size([1, 89, 4096]) -> torch.Size([1, 1, 89, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,256B, BPFP=0.8125 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 22,404B, BPFP=1.9666 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,168B, BPFP=1.4192 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,504B, BPFP=2.1510 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,520B, BPFP=1.6257 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,236B, BPFP=2.2152 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 18,848B, BPFP=1.6545 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,024B, BPFP=2.1966 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,176B, BPFP=1.4199 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,364B, BPFP=2.2265 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 71,460B, BPFP=1.5682 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 89, 128]) - layer.0.v_cache: torch.Size([1, 8, 89, 128]) - layer.1.k_cache: torch.Size([1, 8, 89, 128]) - layer.1.v_cache: torch.Size([1, 8, 89, 128]) - layer.2.k_cache: torch.Size([1, 8, 89, 128]) - layer.2.v_cache: torch.Size([1, 8, 89, 128]) - layer.3.k_cache: torch.Size([1, 8, 89, 128]) - layer.3.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.k_cache: torch.Size([1, 8, 89, 128]) - layer.4.v_cache: torch.Size([1, 8, 89, 128]) - layer.4.output: torch.Size([1, 89, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02424761 8.42342900 - layer.0.v_cache 0.00000029 0.00023341 - layer.1.k_cache 0.00343373 0.81463152 - layer.1.v_cache 0.00000078 0.00085516 - layer.2.k_cache 0.00114421 0.40343150 - layer.2.v_cache 0.00000115 0.00118134 - layer.3.k_cache 0.00135741 0.45146398 - layer.3.v_cache 0.00000236 0.00191754 - layer.4.k_cache 0.00332665 0.83007812 - layer.4.v_cache 0.00000319 0.00317252 - layer.4.output 0.00018046 0.06828655 - ------------------------------------------------------------------------------------- - TOTAL 0.00244566 0.80025288 - (elements=1,275,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1275904 -Total Bytes 272960 -BPFP 1.7115 bits/point -EBPFP 3.4230 equivalent bits/point -MSE 0.800253 ----------------------- -------------------------------------------------------- -Time: 0.512s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 89, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 89, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8003 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample263-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample263-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample274-layer4-item1.zst (42/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample274-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 92, 128) -Output shape: (1, 92, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) -> torch.Size([1, 1, 92, 1024]) - layer.4.output: torch.Size([1, 92, 4096]) -> torch.Size([1, 1, 92, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,980B, BPFP=0.7626 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 22,988B, BPFP=1.9521 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 17,012B, BPFP=1.4446 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,796B, BPFP=2.1056 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,200B, BPFP=1.6304 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,392B, BPFP=2.1562 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,676B, BPFP=1.6709 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,148B, BPFP=2.1355 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,952B, BPFP=1.4395 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,896B, BPFP=2.1990 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 71,604B, BPFP=1.5201 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 92, 128]) - layer.0.v_cache: torch.Size([1, 8, 92, 128]) - layer.1.k_cache: torch.Size([1, 8, 92, 128]) - layer.1.v_cache: torch.Size([1, 8, 92, 128]) - layer.2.k_cache: torch.Size([1, 8, 92, 128]) - layer.2.v_cache: torch.Size([1, 8, 92, 128]) - layer.3.k_cache: torch.Size([1, 8, 92, 128]) - layer.3.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.k_cache: torch.Size([1, 8, 92, 128]) - layer.4.v_cache: torch.Size([1, 8, 92, 128]) - layer.4.output: torch.Size([1, 92, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02551763 9.12625122 - layer.0.v_cache 0.00000027 0.00022675 - layer.1.k_cache 0.00347722 0.83903702 - layer.1.v_cache 0.00000076 0.00075451 - layer.2.k_cache 0.00117662 0.41320158 - layer.2.v_cache 0.00000108 0.00112026 - layer.3.k_cache 0.00141157 0.47226777 - layer.3.v_cache 0.00000195 0.00171487 - layer.4.k_cache 0.00338827 0.90268218 - layer.4.v_cache 0.00000294 0.00305276 - layer.4.output 0.00021021 0.07408235 - ------------------------------------------------------------------------------------- - TOTAL 0.00255851 0.86118845 - (elements=1,318,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1318912 -Total Bytes 277644 -BPFP 1.6841 bits/point -EBPFP 3.3682 equivalent bits/point -MSE 0.861188 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 92, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 92, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8612 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample274-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample274-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample282-layer4-item1.zst (43/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample282-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 87, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 87, 128) -Output shape: (1, 87, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.0.v_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.1.k_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.1.v_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.2.k_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.2.v_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.3.k_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.3.v_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.4.k_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.4.v_cache: torch.Size([1, 8, 87, 128]) -> torch.Size([1, 1, 87, 1024]) - layer.4.output: torch.Size([1, 87, 4096]) -> torch.Size([1, 1, 87, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,036B, BPFP=0.8114 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 22,032B, BPFP=1.9784 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,780B, BPFP=1.4170 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 23,868B, BPFP=2.1433 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,004B, BPFP=1.6167 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,856B, BPFP=2.2320 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 18,520B, BPFP=1.6631 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,512B, BPFP=2.2011 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,844B, BPFP=1.4228 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,032B, BPFP=2.2478 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 69,496B, BPFP=1.5602 -⌛️ [2/4] FRONTEND: Frontend time: 0.208s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 87, 128]) - layer.0.v_cache: torch.Size([1, 8, 87, 128]) - layer.1.k_cache: torch.Size([1, 8, 87, 128]) - layer.1.v_cache: torch.Size([1, 8, 87, 128]) - layer.2.k_cache: torch.Size([1, 8, 87, 128]) - layer.2.v_cache: torch.Size([1, 8, 87, 128]) - layer.3.k_cache: torch.Size([1, 8, 87, 128]) - layer.3.v_cache: torch.Size([1, 8, 87, 128]) - layer.4.k_cache: torch.Size([1, 8, 87, 128]) - layer.4.v_cache: torch.Size([1, 8, 87, 128]) - layer.4.output: torch.Size([1, 87, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.298s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 87, 128]) - layer.0.v_cache: torch.Size([1, 8, 87, 128]) - layer.1.k_cache: torch.Size([1, 8, 87, 128]) - layer.1.v_cache: torch.Size([1, 8, 87, 128]) - layer.2.k_cache: torch.Size([1, 8, 87, 128]) - layer.2.v_cache: torch.Size([1, 8, 87, 128]) - layer.3.k_cache: torch.Size([1, 8, 87, 128]) - layer.3.v_cache: torch.Size([1, 8, 87, 128]) - layer.4.k_cache: torch.Size([1, 8, 87, 128]) - layer.4.v_cache: torch.Size([1, 8, 87, 128]) - layer.4.output: torch.Size([1, 87, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02591714 8.23612678 - layer.0.v_cache 0.00000026 0.00021841 - layer.1.k_cache 0.00374749 0.81172794 - layer.1.v_cache 0.00000077 0.00081451 - layer.2.k_cache 0.00116905 0.40740594 - layer.2.v_cache 0.00000114 0.00115356 - layer.3.k_cache 0.00135864 0.48070272 - layer.3.v_cache 0.00000233 0.00196790 - layer.4.k_cache 0.00339113 0.89602687 - layer.4.v_cache 0.00000304 0.00311895 - layer.4.output 0.00030438 0.08182539 - ------------------------------------------------------------------------------------- - TOTAL 0.00262918 0.79761179 - (elements=1,247,232) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1247232 -Total Bytes 266980 -BPFP 1.7125 bits/point -EBPFP 3.4249 equivalent bits/point -MSE 0.797612 ----------------------- -------------------------------------------------------- -Time: 0.513s Load: 0.007s, Pack+Encode: 0.208s, Decode+Unpack: 0.298s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 87, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 87, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7976 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample282-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample282-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample290-layer4-item1.zst (44/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample290-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 94, 128) -Output shape: (1, 94, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.output: torch.Size([1, 94, 4096]) -> torch.Size([1, 1, 94, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,076B, BPFP=0.7543 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,148B, BPFP=1.9239 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,904B, BPFP=1.4049 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,128B, BPFP=2.0884 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,164B, BPFP=1.5928 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,792B, BPFP=2.1436 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,768B, BPFP=1.6430 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,576B, BPFP=2.1257 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,092B, BPFP=1.4205 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,088B, BPFP=2.1682 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,396B, BPFP=1.5250 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.303s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02423150 8.98378737 - layer.0.v_cache 0.00000028 0.00022820 - layer.1.k_cache 0.00344140 0.84882541 - layer.1.v_cache 0.00000074 0.00080344 - layer.2.k_cache 0.00110852 0.40104570 - layer.2.v_cache 0.00000113 0.00113266 - layer.3.k_cache 0.00135732 0.47470340 - layer.3.v_cache 0.00000223 0.00187131 - layer.4.k_cache 0.00337875 0.82342821 - layer.4.v_cache 0.00000288 0.00305864 - layer.4.output 0.00018208 0.07374552 - ------------------------------------------------------------------------------------- - TOTAL 0.00244665 0.84527618 - (elements=1,347,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1347584 -Total Bytes 281132 -BPFP 1.6690 bits/point -EBPFP 3.3379 equivalent bits/point -MSE 0.845276 ----------------------- -------------------------------------------------------- -Time: 0.519s Load: 0.006s, Pack+Encode: 0.209s, Decode+Unpack: 0.303s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8453 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample290-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample290-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample3-layer4-item1.zst (45/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample3-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 149, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.009s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 149, 128) -Output shape: (1, 149, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.0.v_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.1.k_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.1.v_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.2.k_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.2.v_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.3.k_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.3.v_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.4.k_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.4.v_cache: torch.Size([1, 8, 149, 128]) -> torch.Size([1, 1, 149, 1024]) - layer.4.output: torch.Size([1, 149, 4096]) -> torch.Size([1, 1, 149, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 14,904B, BPFP=0.7815 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 37,976B, BPFP=1.9912 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 26,316B, BPFP=1.3798 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 40,832B, BPFP=2.1409 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 29,740B, BPFP=1.5594 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 41,292B, BPFP=2.1651 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 30,660B, BPFP=1.6076 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 40,628B, BPFP=2.1302 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 26,828B, BPFP=1.4067 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 41,856B, BPFP=2.1946 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 119,580B, BPFP=1.5675 -⌛️ [2/4] FRONTEND: Frontend time: 0.259s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 149, 128]) - layer.0.v_cache: torch.Size([1, 8, 149, 128]) - layer.1.k_cache: torch.Size([1, 8, 149, 128]) - layer.1.v_cache: torch.Size([1, 8, 149, 128]) - layer.2.k_cache: torch.Size([1, 8, 149, 128]) - layer.2.v_cache: torch.Size([1, 8, 149, 128]) - layer.3.k_cache: torch.Size([1, 8, 149, 128]) - layer.3.v_cache: torch.Size([1, 8, 149, 128]) - layer.4.k_cache: torch.Size([1, 8, 149, 128]) - layer.4.v_cache: torch.Size([1, 8, 149, 128]) - layer.4.output: torch.Size([1, 149, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.399s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 149, 128]) - layer.0.v_cache: torch.Size([1, 8, 149, 128]) - layer.1.k_cache: torch.Size([1, 8, 149, 128]) - layer.1.v_cache: torch.Size([1, 8, 149, 128]) - layer.2.k_cache: torch.Size([1, 8, 149, 128]) - layer.2.v_cache: torch.Size([1, 8, 149, 128]) - layer.3.k_cache: torch.Size([1, 8, 149, 128]) - layer.3.v_cache: torch.Size([1, 8, 149, 128]) - layer.4.k_cache: torch.Size([1, 8, 149, 128]) - layer.4.v_cache: torch.Size([1, 8, 149, 128]) - layer.4.output: torch.Size([1, 149, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02473939 7.75279205 - layer.0.v_cache 0.00000028 0.00023695 - layer.1.k_cache 0.00314236 0.75206731 - layer.1.v_cache 0.00000089 0.00085486 - layer.2.k_cache 0.00115430 0.39911355 - layer.2.v_cache 0.00000109 0.00115952 - layer.3.k_cache 0.00134690 0.45526323 - layer.3.v_cache 0.00000207 0.00188995 - layer.4.k_cache 0.00351049 0.81148109 - layer.4.v_cache 0.00000318 0.00327804 - layer.4.output 0.00015300 0.06127328 - ------------------------------------------------------------------------------------- - TOTAL 0.00246521 0.74451640 - (elements=2,136,064) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2136064 -Total Bytes 450612 -BPFP 1.6876 bits/point -EBPFP 3.3753 equivalent bits/point -MSE 0.744516 ----------------------- -------------------------------------------------------- -Time: 0.668s Load: 0.009s, Pack+Encode: 0.259s, Decode+Unpack: 0.399s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 149, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 149, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7445 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample3-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample3-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample307-layer4-item1.zst (46/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample307-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 93, 128) -Output shape: (1, 93, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.output: torch.Size([1, 93, 4096]) -> torch.Size([1, 1, 93, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,888B, BPFP=0.7466 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,012B, BPFP=1.9331 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,948B, BPFP=1.4237 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,288B, BPFP=2.1243 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 19,292B, BPFP=1.6206 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,916B, BPFP=2.1771 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,780B, BPFP=1.6616 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,752B, BPFP=2.1633 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,020B, BPFP=1.4298 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,344B, BPFP=2.2130 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 73,936B, BPFP=1.5528 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.295s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02417885 9.32389914 - layer.0.v_cache 0.00000029 0.00022794 - layer.1.k_cache 0.00337012 0.84044278 - layer.1.v_cache 0.00000087 0.00077425 - layer.2.k_cache 0.00111931 0.39042988 - layer.2.v_cache 0.00000117 0.00104697 - layer.3.k_cache 0.00133665 0.46243819 - layer.3.v_cache 0.00000243 0.00177299 - layer.4.k_cache 0.00331438 0.83358543 - layer.4.v_cache 0.00000316 0.00301860 - layer.4.output 0.00021472 0.08589282 - ------------------------------------------------------------------------------------- - TOTAL 0.00244186 0.87151482 - (elements=1,333,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1333248 -Total Bytes 282176 -BPFP 1.6932 bits/point -EBPFP 3.3863 equivalent bits/point -MSE 0.871515 ----------------------- -------------------------------------------------------- -Time: 0.511s Load: 0.006s, Pack+Encode: 0.210s, Decode+Unpack: 0.295s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8715 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample307-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample307-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample313-layer4-item1.zst (47/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample313-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 81, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 81, 128) -Output shape: (1, 81, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.output: torch.Size([1, 81, 4096]) -> torch.Size([1, 1, 81, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,380B, BPFP=0.8083 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,724B, BPFP=1.9988 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,860B, BPFP=1.4333 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,936B, BPFP=2.2122 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,012B, BPFP=1.6408 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,860B, BPFP=2.3013 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,496B, BPFP=1.6875 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,344B, BPFP=2.2515 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,884B, BPFP=1.4356 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 23,780B, BPFP=2.2936 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 66,324B, BPFP=1.5992 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.output: torch.Size([1, 81, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.294s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.output: torch.Size([1, 81, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02598516 9.86071175 - layer.0.v_cache 0.00000026 0.00023358 - layer.1.k_cache 0.00346748 0.87057683 - layer.1.v_cache 0.00000092 0.00085269 - layer.2.k_cache 0.00113221 0.42418416 - layer.2.v_cache 0.00000143 0.00121005 - layer.3.k_cache 0.00133539 0.48254616 - layer.3.v_cache 0.00000229 0.00191496 - layer.4.k_cache 0.00342405 0.91485106 - layer.4.v_cache 0.00000310 0.00329711 - layer.4.output 0.00020696 0.08856262 - ------------------------------------------------------------------------------------- - TOTAL 0.00258429 0.92247349 - (elements=1,161,216) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1161216 -Total Bytes 253600 -BPFP 1.7471 bits/point -EBPFP 3.4943 equivalent bits/point -MSE 0.922473 ----------------------- -------------------------------------------------------- -Time: 0.506s Load: 0.005s, Pack+Encode: 0.207s, Decode+Unpack: 0.294s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 81, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9225 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample313-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample313-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample319-layer4-item1.zst (48/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample319-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 93, 128) -Output shape: (1, 93, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) -> torch.Size([1, 1, 93, 1024]) - layer.4.output: torch.Size([1, 93, 4096]) -> torch.Size([1, 1, 93, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,908B, BPFP=0.7483 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,012B, BPFP=1.9331 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,916B, BPFP=1.4210 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 25,252B, BPFP=2.1213 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,944B, BPFP=1.5914 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,788B, BPFP=2.1663 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,640B, BPFP=1.6499 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,312B, BPFP=2.1263 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,984B, BPFP=1.4267 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,008B, BPFP=2.1848 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 72,140B, BPFP=1.5150 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 93, 128]) - layer.0.v_cache: torch.Size([1, 8, 93, 128]) - layer.1.k_cache: torch.Size([1, 8, 93, 128]) - layer.1.v_cache: torch.Size([1, 8, 93, 128]) - layer.2.k_cache: torch.Size([1, 8, 93, 128]) - layer.2.v_cache: torch.Size([1, 8, 93, 128]) - layer.3.k_cache: torch.Size([1, 8, 93, 128]) - layer.3.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.k_cache: torch.Size([1, 8, 93, 128]) - layer.4.v_cache: torch.Size([1, 8, 93, 128]) - layer.4.output: torch.Size([1, 93, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02450277 9.70266560 - layer.0.v_cache 0.00000028 0.00024818 - layer.1.k_cache 0.00336978 0.86484208 - layer.1.v_cache 0.00000078 0.00084516 - layer.2.k_cache 0.00117610 0.42104528 - layer.2.v_cache 0.00000111 0.00117205 - layer.3.k_cache 0.00133121 0.47343080 - layer.3.v_cache 0.00000213 0.00192081 - layer.4.k_cache 0.00346230 0.88960200 - layer.4.v_cache 0.00000315 0.00325851 - layer.4.output 0.00018648 0.08317013 - ------------------------------------------------------------------------------------- - TOTAL 0.00247111 0.90655078 - (elements=1,333,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1333248 -Total Bytes 278904 -BPFP 1.6735 bits/point -EBPFP 3.3471 equivalent bits/point -MSE 0.906551 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 93, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 93, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9066 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample319-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample319-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample333-layer4-item1.zst (49/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample333-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 90, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 90, 128) -Output shape: (1, 90, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.0.v_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.1.k_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.1.v_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.2.k_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.2.v_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.3.k_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.3.v_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.4.k_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.4.v_cache: torch.Size([1, 8, 90, 128]) -> torch.Size([1, 1, 90, 1024]) - layer.4.output: torch.Size([1, 90, 4096]) -> torch.Size([1, 1, 90, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,108B, BPFP=0.7906 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 22,496B, BPFP=1.9528 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,248B, BPFP=1.4104 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,352B, BPFP=2.1139 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,428B, BPFP=1.5997 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,148B, BPFP=2.1830 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,140B, BPFP=1.6615 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 24,880B, BPFP=2.1597 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 16,120B, BPFP=1.3993 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 25,188B, BPFP=2.1865 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 68,892B, BPFP=1.4951 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 90, 128]) - layer.0.v_cache: torch.Size([1, 8, 90, 128]) - layer.1.k_cache: torch.Size([1, 8, 90, 128]) - layer.1.v_cache: torch.Size([1, 8, 90, 128]) - layer.2.k_cache: torch.Size([1, 8, 90, 128]) - layer.2.v_cache: torch.Size([1, 8, 90, 128]) - layer.3.k_cache: torch.Size([1, 8, 90, 128]) - layer.3.v_cache: torch.Size([1, 8, 90, 128]) - layer.4.k_cache: torch.Size([1, 8, 90, 128]) - layer.4.v_cache: torch.Size([1, 8, 90, 128]) - layer.4.output: torch.Size([1, 90, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 90, 128]) - layer.0.v_cache: torch.Size([1, 8, 90, 128]) - layer.1.k_cache: torch.Size([1, 8, 90, 128]) - layer.1.v_cache: torch.Size([1, 8, 90, 128]) - layer.2.k_cache: torch.Size([1, 8, 90, 128]) - layer.2.v_cache: torch.Size([1, 8, 90, 128]) - layer.3.k_cache: torch.Size([1, 8, 90, 128]) - layer.3.v_cache: torch.Size([1, 8, 90, 128]) - layer.4.k_cache: torch.Size([1, 8, 90, 128]) - layer.4.v_cache: torch.Size([1, 8, 90, 128]) - layer.4.output: torch.Size([1, 90, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02906363 8.88964776 - layer.0.v_cache 0.00000028 0.00022947 - layer.1.k_cache 0.00333766 0.83537242 - layer.1.v_cache 0.00000075 0.00074804 - layer.2.k_cache 0.00114903 0.40991643 - layer.2.v_cache 0.00000112 0.00103845 - layer.3.k_cache 0.00131913 0.47667987 - layer.3.v_cache 0.00000216 0.00175300 - layer.4.k_cache 0.00329656 0.87162408 - layer.4.v_cache 0.00000300 0.00291086 - layer.4.output 0.00024019 0.06803242 - ------------------------------------------------------------------------------------- - TOTAL 0.00279529 0.84014643 - (elements=1,290,240) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1290240 -Total Bytes 270000 -BPFP 1.6741 bits/point -EBPFP 3.3482 equivalent bits/point -MSE 0.840146 ----------------------- -------------------------------------------------------- -Time: 0.495s Load: 0.005s, Pack+Encode: 0.205s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 90, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 90, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8401 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample333-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample333-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample36-layer4-item1.zst (50/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample36-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 119, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 119, 128) -Output shape: (1, 119, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.0.v_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.1.k_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.1.v_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.2.k_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.2.v_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.3.k_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.3.v_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.4.k_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.4.v_cache: torch.Size([1, 8, 119, 128]) -> torch.Size([1, 1, 119, 1024]) - layer.4.output: torch.Size([1, 119, 4096]) -> torch.Size([1, 1, 119, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,672B, BPFP=0.7663 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,448B, BPFP=1.9333 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 20,804B, BPFP=1.3658 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,436B, BPFP=2.0638 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,452B, BPFP=1.5397 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,888B, BPFP=2.0935 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,952B, BPFP=1.5725 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 31,320B, BPFP=2.0562 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 21,072B, BPFP=1.3834 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,876B, BPFP=2.0927 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 94,584B, BPFP=1.5524 -⌛️ [2/4] FRONTEND: Frontend time: 0.206s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 119, 128]) - layer.0.v_cache: torch.Size([1, 8, 119, 128]) - layer.1.k_cache: torch.Size([1, 8, 119, 128]) - layer.1.v_cache: torch.Size([1, 8, 119, 128]) - layer.2.k_cache: torch.Size([1, 8, 119, 128]) - layer.2.v_cache: torch.Size([1, 8, 119, 128]) - layer.3.k_cache: torch.Size([1, 8, 119, 128]) - layer.3.v_cache: torch.Size([1, 8, 119, 128]) - layer.4.k_cache: torch.Size([1, 8, 119, 128]) - layer.4.v_cache: torch.Size([1, 8, 119, 128]) - layer.4.output: torch.Size([1, 119, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.289s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 119, 128]) - layer.0.v_cache: torch.Size([1, 8, 119, 128]) - layer.1.k_cache: torch.Size([1, 8, 119, 128]) - layer.1.v_cache: torch.Size([1, 8, 119, 128]) - layer.2.k_cache: torch.Size([1, 8, 119, 128]) - layer.2.v_cache: torch.Size([1, 8, 119, 128]) - layer.3.k_cache: torch.Size([1, 8, 119, 128]) - layer.3.v_cache: torch.Size([1, 8, 119, 128]) - layer.4.k_cache: torch.Size([1, 8, 119, 128]) - layer.4.v_cache: torch.Size([1, 8, 119, 128]) - layer.4.output: torch.Size([1, 119, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02458241 7.56492294 - layer.0.v_cache 0.00000028 0.00023684 - layer.1.k_cache 0.00331476 0.73505876 - layer.1.v_cache 0.00000080 0.00086680 - layer.2.k_cache 0.00114230 0.40829756 - layer.2.v_cache 0.00000115 0.00122776 - layer.3.k_cache 0.00138709 0.46509090 - layer.3.v_cache 0.00000225 0.00198483 - layer.4.k_cache 0.00333808 0.85576546 - layer.4.v_cache 0.00000314 0.00325181 - layer.4.output 0.00020645 0.08369014 - ------------------------------------------------------------------------------------- - TOTAL 0.00247129 0.74081887 - (elements=1,705,984) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1705984 -Total Bytes 351504 -BPFP 1.6483 bits/point -EBPFP 3.2967 equivalent bits/point -MSE 0.740819 ----------------------- -------------------------------------------------------- -Time: 0.502s Load: 0.007s, Pack+Encode: 0.206s, Decode+Unpack: 0.289s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 119, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 119, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7408 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample36-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample36-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample365-layer4-item1.zst (51/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample365-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 82, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 82, 128) -Output shape: (1, 82, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) -> torch.Size([1, 1, 82, 1024]) - layer.4.output: torch.Size([1, 82, 4096]) -> torch.Size([1, 1, 82, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,940B, BPFP=0.7565 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,856B, BPFP=1.9870 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,836B, BPFP=1.4135 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,832B, BPFP=2.1753 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 16,836B, BPFP=1.6040 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,676B, BPFP=2.2557 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,444B, BPFP=1.6620 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,348B, BPFP=2.2245 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,748B, BPFP=1.4051 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 23,684B, BPFP=2.2565 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 65,280B, BPFP=1.5549 -⌛️ [2/4] FRONTEND: Frontend time: 0.206s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.output: torch.Size([1, 82, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.286s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 82, 128]) - layer.0.v_cache: torch.Size([1, 8, 82, 128]) - layer.1.k_cache: torch.Size([1, 8, 82, 128]) - layer.1.v_cache: torch.Size([1, 8, 82, 128]) - layer.2.k_cache: torch.Size([1, 8, 82, 128]) - layer.2.v_cache: torch.Size([1, 8, 82, 128]) - layer.3.k_cache: torch.Size([1, 8, 82, 128]) - layer.3.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.k_cache: torch.Size([1, 8, 82, 128]) - layer.4.v_cache: torch.Size([1, 8, 82, 128]) - layer.4.output: torch.Size([1, 82, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02610582 9.99716559 - layer.0.v_cache 0.00000027 0.00023422 - layer.1.k_cache 0.00344214 0.88998506 - layer.1.v_cache 0.00000078 0.00081307 - layer.2.k_cache 0.00114699 0.42041592 - layer.2.v_cache 0.00000119 0.00116604 - layer.3.k_cache 0.00140111 0.47612967 - layer.3.v_cache 0.00000200 0.00185542 - layer.4.k_cache 0.00332288 0.89503312 - layer.4.v_cache 0.00000299 0.00303389 - layer.4.output 0.00018752 0.08851091 - ------------------------------------------------------------------------------------- - TOTAL 0.00258402 0.93141969 - (elements=1,175,552) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1175552 -Total Bytes 251480 -BPFP 1.7114 bits/point -EBPFP 3.4228 equivalent bits/point -MSE 0.931420 ----------------------- -------------------------------------------------------- -Time: 0.498s Load: 0.006s, Pack+Encode: 0.206s, Decode+Unpack: 0.286s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 82, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 82, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9314 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample365-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample365-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample38-layer4-item1.zst (52/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample38-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 124, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 124, 128) -Output shape: (1, 124, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.0.v_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.1.k_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.1.v_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.2.k_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.2.v_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.3.k_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.3.v_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.4.k_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.4.v_cache: torch.Size([1, 8, 124, 128]) -> torch.Size([1, 1, 124, 1024]) - layer.4.output: torch.Size([1, 124, 4096]) -> torch.Size([1, 1, 124, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 12,012B, BPFP=0.7568 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,828B, BPFP=1.8793 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 21,868B, BPFP=1.3778 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,564B, BPFP=1.9887 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 24,180B, BPFP=1.5234 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,836B, BPFP=2.0058 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,900B, BPFP=1.5688 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 31,264B, BPFP=1.9698 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 22,268B, BPFP=1.4030 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,952B, BPFP=2.0131 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 94,260B, BPFP=1.4847 -⌛️ [2/4] FRONTEND: Frontend time: 0.208s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 124, 128]) - layer.0.v_cache: torch.Size([1, 8, 124, 128]) - layer.1.k_cache: torch.Size([1, 8, 124, 128]) - layer.1.v_cache: torch.Size([1, 8, 124, 128]) - layer.2.k_cache: torch.Size([1, 8, 124, 128]) - layer.2.v_cache: torch.Size([1, 8, 124, 128]) - layer.3.k_cache: torch.Size([1, 8, 124, 128]) - layer.3.v_cache: torch.Size([1, 8, 124, 128]) - layer.4.k_cache: torch.Size([1, 8, 124, 128]) - layer.4.v_cache: torch.Size([1, 8, 124, 128]) - layer.4.output: torch.Size([1, 124, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.291s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 124, 128]) - layer.0.v_cache: torch.Size([1, 8, 124, 128]) - layer.1.k_cache: torch.Size([1, 8, 124, 128]) - layer.1.v_cache: torch.Size([1, 8, 124, 128]) - layer.2.k_cache: torch.Size([1, 8, 124, 128]) - layer.2.v_cache: torch.Size([1, 8, 124, 128]) - layer.3.k_cache: torch.Size([1, 8, 124, 128]) - layer.3.v_cache: torch.Size([1, 8, 124, 128]) - layer.4.k_cache: torch.Size([1, 8, 124, 128]) - layer.4.v_cache: torch.Size([1, 8, 124, 128]) - layer.4.output: torch.Size([1, 124, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02574280 7.66323360 - layer.0.v_cache 0.00000027 0.00022944 - layer.1.k_cache 0.00324291 0.72658859 - layer.1.v_cache 0.00000080 0.00078557 - layer.2.k_cache 0.00117169 0.39671858 - layer.2.v_cache 0.00000108 0.00109450 - layer.3.k_cache 0.00137525 0.44905552 - layer.3.v_cache 0.00000203 0.00175134 - layer.4.k_cache 0.00340243 0.79833172 - layer.4.v_cache 0.00000299 0.00301618 - layer.4.output 0.00019780 0.07330540 - ------------------------------------------------------------------------------------- - TOTAL 0.00255239 0.73814476 - (elements=1,777,664) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1777664 -Total Bytes 355932 -BPFP 1.6018 bits/point -EBPFP 3.2036 equivalent bits/point -MSE 0.738145 ----------------------- -------------------------------------------------------- -Time: 0.507s Load: 0.008s, Pack+Encode: 0.208s, Decode+Unpack: 0.291s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 124, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 124, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7381 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample38-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample38-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample388-layer4-item1.zst (53/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample388-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 85, 128) -Output shape: (1, 85, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.output: torch.Size([1, 85, 4096]) -> torch.Size([1, 1, 85, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,200B, BPFP=0.7537 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,000B, BPFP=1.9301 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,076B, BPFP=1.3857 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 23,160B, BPFP=2.1287 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,264B, BPFP=1.5868 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,940B, BPFP=2.2004 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,932B, BPFP=1.6482 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,728B, BPFP=2.1809 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,364B, BPFP=1.4121 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 24,132B, BPFP=2.2180 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 66,768B, BPFP=1.5342 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.286s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02690007 10.40428682 - layer.0.v_cache 0.00000028 0.00024701 - layer.1.k_cache 0.00337091 0.87022831 - layer.1.v_cache 0.00000080 0.00079869 - layer.2.k_cache 0.00114287 0.42341470 - layer.2.v_cache 0.00000105 0.00108713 - layer.3.k_cache 0.00139664 0.49459929 - layer.3.v_cache 0.00000196 0.00174345 - layer.4.k_cache 0.00332559 0.91407058 - layer.4.v_cache 0.00000294 0.00305859 - layer.4.output 0.00025832 0.09658302 - ------------------------------------------------------------------------------------- - TOTAL 0.00265546 0.96427619 - (elements=1,218,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1218560 -Total Bytes 256564 -BPFP 1.6844 bits/point -EBPFP 3.3687 equivalent bits/point -MSE 0.964276 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.006s, Pack+Encode: 0.203s, Decode+Unpack: 0.286s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9643 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample388-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample388-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample390-layer4-item1.zst (54/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample390-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 74, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 74, 128) -Output shape: (1, 74, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.output: torch.Size([1, 74, 4096]) -> torch.Size([1, 1, 74, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,720B, BPFP=0.8150 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 19,548B, BPFP=2.0638 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,776B, BPFP=1.4544 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 21,284B, BPFP=2.2470 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,272B, BPFP=1.6123 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,680B, BPFP=2.2889 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,944B, BPFP=1.6833 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 21,320B, BPFP=2.2508 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,512B, BPFP=1.4265 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 21,836B, BPFP=2.3053 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 59,868B, BPFP=1.5801 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.output: torch.Size([1, 74, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.output: torch.Size([1, 74, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02896725 10.41281375 - layer.0.v_cache 0.00000027 0.00024338 - layer.1.k_cache 0.00342263 0.92598291 - layer.1.v_cache 0.00000079 0.00087832 - layer.2.k_cache 0.00118243 0.44673069 - layer.2.v_cache 0.00000111 0.00117136 - layer.3.k_cache 0.00141981 0.51034283 - layer.3.v_cache 0.00000233 0.00190041 - layer.4.k_cache 0.00326747 0.94969383 - layer.4.v_cache 0.00000299 0.00311502 - layer.4.output 0.00023783 0.09733059 - ------------------------------------------------------------------------------------- - TOTAL 0.00280131 0.97444249 - (elements=1,060,864) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1060864 -Total Bytes 231760 -BPFP 1.7477 bits/point -EBPFP 3.4954 equivalent bits/point -MSE 0.974442 ----------------------- -------------------------------------------------------- -Time: 0.495s Load: 0.005s, Pack+Encode: 0.204s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 74, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9744 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample390-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample390-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample40-layer4-item1.zst (55/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample40-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 126, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 126, 128) -Output shape: (1, 126, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.0.v_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.1.k_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.1.v_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.2.k_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.2.v_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.3.k_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.3.v_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.4.k_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.4.v_cache: torch.Size([1, 8, 126, 128]) -> torch.Size([1, 1, 126, 1024]) - layer.4.output: torch.Size([1, 126, 4096]) -> torch.Size([1, 1, 126, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 12,744B, BPFP=0.7902 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,396B, BPFP=1.8227 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 21,840B, BPFP=1.3542 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,044B, BPFP=1.9249 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,888B, BPFP=1.4812 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,264B, BPFP=1.9385 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,636B, BPFP=1.5275 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,848B, BPFP=1.9127 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 22,260B, BPFP=1.3802 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,508B, BPFP=1.9536 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 92,676B, BPFP=1.4366 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 126, 128]) - layer.0.v_cache: torch.Size([1, 8, 126, 128]) - layer.1.k_cache: torch.Size([1, 8, 126, 128]) - layer.1.v_cache: torch.Size([1, 8, 126, 128]) - layer.2.k_cache: torch.Size([1, 8, 126, 128]) - layer.2.v_cache: torch.Size([1, 8, 126, 128]) - layer.3.k_cache: torch.Size([1, 8, 126, 128]) - layer.3.v_cache: torch.Size([1, 8, 126, 128]) - layer.4.k_cache: torch.Size([1, 8, 126, 128]) - layer.4.v_cache: torch.Size([1, 8, 126, 128]) - layer.4.output: torch.Size([1, 126, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.295s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 126, 128]) - layer.0.v_cache: torch.Size([1, 8, 126, 128]) - layer.1.k_cache: torch.Size([1, 8, 126, 128]) - layer.1.v_cache: torch.Size([1, 8, 126, 128]) - layer.2.k_cache: torch.Size([1, 8, 126, 128]) - layer.2.v_cache: torch.Size([1, 8, 126, 128]) - layer.3.k_cache: torch.Size([1, 8, 126, 128]) - layer.3.v_cache: torch.Size([1, 8, 126, 128]) - layer.4.k_cache: torch.Size([1, 8, 126, 128]) - layer.4.v_cache: torch.Size([1, 8, 126, 128]) - layer.4.output: torch.Size([1, 126, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02675629 7.95686801 - layer.0.v_cache 0.00000026 0.00021757 - layer.1.k_cache 0.00313964 0.78679633 - layer.1.v_cache 0.00000077 0.00077877 - layer.2.k_cache 0.00115655 0.41648329 - layer.2.v_cache 0.00000107 0.00112086 - layer.3.k_cache 0.00144242 0.46812799 - layer.3.v_cache 0.00000205 0.00181025 - layer.4.k_cache 0.00340494 0.89688353 - layer.4.v_cache 0.00000295 0.00309948 - layer.4.output 0.00023638 0.07830836 - ------------------------------------------------------------------------------------- - TOTAL 0.00263232 0.77467282 - (elements=1,806,336) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1806336 -Total Bytes 352104 -BPFP 1.5594 bits/point -EBPFP 3.1188 equivalent bits/point -MSE 0.774673 ----------------------- -------------------------------------------------------- -Time: 0.511s Load: 0.008s, Pack+Encode: 0.207s, Decode+Unpack: 0.295s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 126, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 126, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7747 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample40-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample40-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample412-layer4-item1.zst (56/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample412-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 76, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 76, 128) -Output shape: (1, 76, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.0.v_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.1.k_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.1.v_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.2.k_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.2.v_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.3.k_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.3.v_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.4.k_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.4.v_cache: torch.Size([1, 8, 76, 128]) -> torch.Size([1, 1, 76, 1024]) - layer.4.output: torch.Size([1, 76, 4096]) -> torch.Size([1, 1, 76, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,532B, BPFP=0.7743 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 19,232B, BPFP=1.9770 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,796B, BPFP=1.4182 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 21,212B, BPFP=2.1805 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,528B, BPFP=1.5962 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,836B, BPFP=2.2447 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 16,164B, BPFP=1.6616 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 21,688B, BPFP=2.2294 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,744B, BPFP=1.4128 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 21,976B, BPFP=2.2590 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 60,764B, BPFP=1.5616 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 76, 128]) - layer.0.v_cache: torch.Size([1, 8, 76, 128]) - layer.1.k_cache: torch.Size([1, 8, 76, 128]) - layer.1.v_cache: torch.Size([1, 8, 76, 128]) - layer.2.k_cache: torch.Size([1, 8, 76, 128]) - layer.2.v_cache: torch.Size([1, 8, 76, 128]) - layer.3.k_cache: torch.Size([1, 8, 76, 128]) - layer.3.v_cache: torch.Size([1, 8, 76, 128]) - layer.4.k_cache: torch.Size([1, 8, 76, 128]) - layer.4.v_cache: torch.Size([1, 8, 76, 128]) - layer.4.output: torch.Size([1, 76, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.286s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 76, 128]) - layer.0.v_cache: torch.Size([1, 8, 76, 128]) - layer.1.k_cache: torch.Size([1, 8, 76, 128]) - layer.1.v_cache: torch.Size([1, 8, 76, 128]) - layer.2.k_cache: torch.Size([1, 8, 76, 128]) - layer.2.v_cache: torch.Size([1, 8, 76, 128]) - layer.3.k_cache: torch.Size([1, 8, 76, 128]) - layer.3.v_cache: torch.Size([1, 8, 76, 128]) - layer.4.k_cache: torch.Size([1, 8, 76, 128]) - layer.4.v_cache: torch.Size([1, 8, 76, 128]) - layer.4.output: torch.Size([1, 76, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02706527 10.21134226 - layer.0.v_cache 0.00000027 0.00022884 - layer.1.k_cache 0.00360181 0.93624085 - layer.1.v_cache 0.00000073 0.00075052 - layer.2.k_cache 0.00113352 0.44640461 - layer.2.v_cache 0.00000108 0.00113537 - layer.3.k_cache 0.00139474 0.50229233 - layer.3.v_cache 0.00000224 0.00182678 - layer.4.k_cache 0.00318908 0.95822003 - layer.4.v_cache 0.00000296 0.00302649 - layer.4.output 0.00022408 0.10045953 - ------------------------------------------------------------------------------------- - TOTAL 0.00266343 0.96166473 - (elements=1,089,536) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1089536 -Total Bytes 233472 -BPFP 1.7143 bits/point -EBPFP 3.4286 equivalent bits/point -MSE 0.961665 ----------------------- -------------------------------------------------------- -Time: 0.495s Load: 0.005s, Pack+Encode: 0.203s, Decode+Unpack: 0.286s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 76, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 76, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9617 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample412-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample412-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample414-layer4-item1.zst (57/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample414-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 85, 128) -Output shape: (1, 85, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) -> torch.Size([1, 1, 85, 1024]) - layer.4.output: torch.Size([1, 85, 4096]) -> torch.Size([1, 1, 85, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,188B, BPFP=0.7526 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,356B, BPFP=1.9629 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,224B, BPFP=1.3993 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 23,172B, BPFP=2.1298 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,384B, BPFP=1.5978 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,072B, BPFP=2.2125 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,920B, BPFP=1.6471 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,928B, BPFP=2.1993 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,348B, BPFP=1.4107 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 24,248B, BPFP=2.2287 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 67,568B, BPFP=1.5526 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.290s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 85, 128]) - layer.0.v_cache: torch.Size([1, 8, 85, 128]) - layer.1.k_cache: torch.Size([1, 8, 85, 128]) - layer.1.v_cache: torch.Size([1, 8, 85, 128]) - layer.2.k_cache: torch.Size([1, 8, 85, 128]) - layer.2.v_cache: torch.Size([1, 8, 85, 128]) - layer.3.k_cache: torch.Size([1, 8, 85, 128]) - layer.3.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.k_cache: torch.Size([1, 8, 85, 128]) - layer.4.v_cache: torch.Size([1, 8, 85, 128]) - layer.4.output: torch.Size([1, 85, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02461138 9.58130314 - layer.0.v_cache 0.00000028 0.00022038 - layer.1.k_cache 0.00340719 0.88068982 - layer.1.v_cache 0.00000073 0.00070527 - layer.2.k_cache 0.00115796 0.41469017 - layer.2.v_cache 0.00000109 0.00103036 - layer.3.k_cache 0.00135314 0.47658216 - layer.3.v_cache 0.00000196 0.00161463 - layer.4.k_cache 0.00328730 0.84037099 - layer.4.v_cache 0.00000286 0.00285482 - layer.4.output 0.00022769 0.08982837 - ------------------------------------------------------------------------------------- - TOTAL 0.00248105 0.89709823 - (elements=1,218,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1218560 -Total Bytes 258408 -BPFP 1.6965 bits/point -EBPFP 3.3930 equivalent bits/point -MSE 0.897098 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.006s, Pack+Encode: 0.205s, Decode+Unpack: 0.290s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 85, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 85, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8971 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample414-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample414-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample42-layer4-item1.zst (58/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample42-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 134, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 134, 128) -Output shape: (1, 134, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.0.v_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.1.k_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.1.v_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.2.k_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.2.v_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.3.k_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.3.v_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.4.k_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.4.v_cache: torch.Size([1, 8, 134, 128]) -> torch.Size([1, 1, 134, 1024]) - layer.4.output: torch.Size([1, 134, 4096]) -> torch.Size([1, 1, 134, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 13,392B, BPFP=0.7808 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 34,224B, BPFP=1.9953 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 24,188B, BPFP=1.4102 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 36,872B, BPFP=2.1497 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 27,248B, BPFP=1.5886 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 37,264B, BPFP=2.1726 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 28,640B, BPFP=1.6698 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 37,148B, BPFP=2.1658 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 24,820B, BPFP=1.4471 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 37,956B, BPFP=2.2129 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 103,736B, BPFP=1.5120 -⌛️ [2/4] FRONTEND: Frontend time: 0.252s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 134, 128]) - layer.0.v_cache: torch.Size([1, 8, 134, 128]) - layer.1.k_cache: torch.Size([1, 8, 134, 128]) - layer.1.v_cache: torch.Size([1, 8, 134, 128]) - layer.2.k_cache: torch.Size([1, 8, 134, 128]) - layer.2.v_cache: torch.Size([1, 8, 134, 128]) - layer.3.k_cache: torch.Size([1, 8, 134, 128]) - layer.3.v_cache: torch.Size([1, 8, 134, 128]) - layer.4.k_cache: torch.Size([1, 8, 134, 128]) - layer.4.v_cache: torch.Size([1, 8, 134, 128]) - layer.4.output: torch.Size([1, 134, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.388s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 134, 128]) - layer.0.v_cache: torch.Size([1, 8, 134, 128]) - layer.1.k_cache: torch.Size([1, 8, 134, 128]) - layer.1.v_cache: torch.Size([1, 8, 134, 128]) - layer.2.k_cache: torch.Size([1, 8, 134, 128]) - layer.2.v_cache: torch.Size([1, 8, 134, 128]) - layer.3.k_cache: torch.Size([1, 8, 134, 128]) - layer.3.v_cache: torch.Size([1, 8, 134, 128]) - layer.4.k_cache: torch.Size([1, 8, 134, 128]) - layer.4.v_cache: torch.Size([1, 8, 134, 128]) - layer.4.output: torch.Size([1, 134, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02623714 8.99444489 - layer.0.v_cache 0.00000029 0.00022976 - layer.1.k_cache 0.00318200 0.78211121 - layer.1.v_cache 0.00000078 0.00076211 - layer.2.k_cache 0.00115661 0.40110130 - layer.2.v_cache 0.00000102 0.00099686 - layer.3.k_cache 0.00142778 0.47199167 - layer.3.v_cache 0.00000209 0.00175907 - layer.4.k_cache 0.00344207 0.88896168 - layer.4.v_cache 0.00000292 0.00291355 - layer.4.output 0.00019652 0.07342665 - ------------------------------------------------------------------------------------- - TOTAL 0.00258848 0.84564133 - (elements=1,921,024) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1921024 -Total Bytes 405488 -BPFP 1.6886 bits/point -EBPFP 3.3773 equivalent bits/point -MSE 0.845641 ----------------------- -------------------------------------------------------- -Time: 0.647s Load: 0.007s, Pack+Encode: 0.252s, Decode+Unpack: 0.388s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 134, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 134, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8456 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample42-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample42-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample44-layer4-item1.zst (59/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample44-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 132, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 132, 128) -Output shape: (1, 132, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.0.v_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.1.k_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.1.v_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.2.k_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.2.v_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.3.k_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.3.v_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.4.k_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.4.v_cache: torch.Size([1, 8, 132, 128]) -> torch.Size([1, 1, 132, 1024]) - layer.4.output: torch.Size([1, 132, 4096]) -> torch.Size([1, 1, 132, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 13,928B, BPFP=0.8243 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 34,104B, BPFP=2.0185 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 24,012B, BPFP=1.4212 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 36,916B, BPFP=2.1849 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 26,944B, BPFP=1.5947 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 37,616B, BPFP=2.2263 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 28,272B, BPFP=1.6733 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 37,052B, BPFP=2.1929 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 24,628B, BPFP=1.4576 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 37,736B, BPFP=2.2334 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 107,084B, BPFP=1.5845 -⌛️ [2/4] FRONTEND: Frontend time: 0.254s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 132, 128]) - layer.0.v_cache: torch.Size([1, 8, 132, 128]) - layer.1.k_cache: torch.Size([1, 8, 132, 128]) - layer.1.v_cache: torch.Size([1, 8, 132, 128]) - layer.2.k_cache: torch.Size([1, 8, 132, 128]) - layer.2.v_cache: torch.Size([1, 8, 132, 128]) - layer.3.k_cache: torch.Size([1, 8, 132, 128]) - layer.3.v_cache: torch.Size([1, 8, 132, 128]) - layer.4.k_cache: torch.Size([1, 8, 132, 128]) - layer.4.v_cache: torch.Size([1, 8, 132, 128]) - layer.4.output: torch.Size([1, 132, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.391s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 132, 128]) - layer.0.v_cache: torch.Size([1, 8, 132, 128]) - layer.1.k_cache: torch.Size([1, 8, 132, 128]) - layer.1.v_cache: torch.Size([1, 8, 132, 128]) - layer.2.k_cache: torch.Size([1, 8, 132, 128]) - layer.2.v_cache: torch.Size([1, 8, 132, 128]) - layer.3.k_cache: torch.Size([1, 8, 132, 128]) - layer.3.v_cache: torch.Size([1, 8, 132, 128]) - layer.4.k_cache: torch.Size([1, 8, 132, 128]) - layer.4.v_cache: torch.Size([1, 8, 132, 128]) - layer.4.output: torch.Size([1, 132, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02542288 8.89218787 - layer.0.v_cache 0.00000027 0.00023348 - layer.1.k_cache 0.00323066 0.79840724 - layer.1.v_cache 0.00000090 0.00082314 - layer.2.k_cache 0.00115984 0.41324673 - layer.2.v_cache 0.00000123 0.00115243 - layer.3.k_cache 0.00138390 0.48339717 - layer.3.v_cache 0.00000215 0.00191770 - layer.4.k_cache 0.00364779 0.85500307 - layer.4.v_cache 0.00000297 0.00296955 - layer.4.output 0.00020533 0.08450513 - ------------------------------------------------------------------------------------- - TOTAL 0.00254814 0.84195421 - (elements=1,892,352) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1892352 -Total Bytes 408292 -BPFP 1.7261 bits/point -EBPFP 3.4521 equivalent bits/point -MSE 0.841954 ----------------------- -------------------------------------------------------- -Time: 0.653s Load: 0.007s, Pack+Encode: 0.254s, Decode+Unpack: 0.391s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 132, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 132, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8420 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample44-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample44-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample454-layer4-item1.zst (60/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample454-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 77, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 77, 128) -Output shape: (1, 77, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.output: torch.Size([1, 77, 4096]) -> torch.Size([1, 1, 77, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,124B, BPFP=0.8243 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,160B, BPFP=2.0455 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,120B, BPFP=1.4326 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,048B, BPFP=2.2370 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,992B, BPFP=1.6226 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 22,772B, BPFP=2.3105 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 16,460B, BPFP=1.6700 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 22,292B, BPFP=2.2618 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,088B, BPFP=1.4294 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 22,724B, BPFP=2.3056 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 64,064B, BPFP=1.6250 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.output: torch.Size([1, 77, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.287s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.output: torch.Size([1, 77, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02625101 9.57146503 - layer.0.v_cache 0.00000028 0.00024757 - layer.1.k_cache 0.00355533 0.94241353 - layer.1.v_cache 0.00000093 0.00084309 - layer.2.k_cache 0.00115642 0.43220986 - layer.2.v_cache 0.00000119 0.00118649 - layer.3.k_cache 0.00136722 0.50305443 - layer.3.v_cache 0.00000230 0.00197534 - layer.4.k_cache 0.00337541 0.91250264 - layer.4.v_cache 0.00000304 0.00325855 - layer.4.output 0.00023573 0.09894578 - ------------------------------------------------------------------------------------- - TOTAL 0.00261829 0.91178141 - (elements=1,103,872) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1103872 -Total Bytes 242844 -BPFP 1.7599 bits/point -EBPFP 3.5199 equivalent bits/point -MSE 0.911781 ----------------------- -------------------------------------------------------- -Time: 0.495s Load: 0.005s, Pack+Encode: 0.204s, Decode+Unpack: 0.287s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 77, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9118 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample454-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample454-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample464-layer4-item1.zst (61/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample464-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 94, 128) -Output shape: (1, 94, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) -> torch.Size([1, 1, 94, 1024]) - layer.4.output: torch.Size([1, 94, 4096]) -> torch.Size([1, 1, 94, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 9,236B, BPFP=0.7676 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 23,304B, BPFP=1.9368 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 16,732B, BPFP=1.3906 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 24,712B, BPFP=2.0539 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 18,880B, BPFP=1.5691 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 25,540B, BPFP=2.1227 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 19,916B, BPFP=1.6553 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 25,488B, BPFP=2.1184 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 17,120B, BPFP=1.4229 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 26,064B, BPFP=2.1662 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 74,048B, BPFP=1.5386 -⌛️ [2/4] FRONTEND: Frontend time: 0.202s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.287s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 94, 128]) - layer.0.v_cache: torch.Size([1, 8, 94, 128]) - layer.1.k_cache: torch.Size([1, 8, 94, 128]) - layer.1.v_cache: torch.Size([1, 8, 94, 128]) - layer.2.k_cache: torch.Size([1, 8, 94, 128]) - layer.2.v_cache: torch.Size([1, 8, 94, 128]) - layer.3.k_cache: torch.Size([1, 8, 94, 128]) - layer.3.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.k_cache: torch.Size([1, 8, 94, 128]) - layer.4.v_cache: torch.Size([1, 8, 94, 128]) - layer.4.output: torch.Size([1, 94, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02258782 8.86936529 - layer.0.v_cache 0.00000029 0.00023455 - layer.1.k_cache 0.00339276 0.85268094 - layer.1.v_cache 0.00000074 0.00069562 - layer.2.k_cache 0.00116230 0.38564000 - layer.2.v_cache 0.00000115 0.00104172 - layer.3.k_cache 0.00135573 0.45952915 - layer.3.v_cache 0.00000225 0.00177230 - layer.4.k_cache 0.00337629 0.82706735 - layer.4.v_cache 0.00000290 0.00269800 - layer.4.output 0.00017695 0.08478082 - ------------------------------------------------------------------------------------- - TOTAL 0.00232786 0.83856059 - (elements=1,347,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1347584 -Total Bytes 281040 -BPFP 1.6684 bits/point -EBPFP 3.3368 equivalent bits/point -MSE 0.838561 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.007s, Pack+Encode: 0.202s, Decode+Unpack: 0.287s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 94, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 94, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8386 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample464-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample464-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample478-layer4-item1.zst (62/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample478-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 77, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 77, 128) -Output shape: (1, 77, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) -> torch.Size([1, 1, 77, 1024]) - layer.4.output: torch.Size([1, 77, 4096]) -> torch.Size([1, 1, 77, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,008B, BPFP=0.8125 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 19,900B, BPFP=2.0191 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,116B, BPFP=1.4322 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 21,876B, BPFP=2.2196 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,920B, BPFP=1.6153 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 22,328B, BPFP=2.2654 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 16,552B, BPFP=1.6794 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 21,976B, BPFP=2.2297 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,040B, BPFP=1.4245 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 22,508B, BPFP=2.2837 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 62,608B, BPFP=1.5881 -⌛️ [2/4] FRONTEND: Frontend time: 0.202s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.output: torch.Size([1, 77, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.284s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 77, 128]) - layer.0.v_cache: torch.Size([1, 8, 77, 128]) - layer.1.k_cache: torch.Size([1, 8, 77, 128]) - layer.1.v_cache: torch.Size([1, 8, 77, 128]) - layer.2.k_cache: torch.Size([1, 8, 77, 128]) - layer.2.v_cache: torch.Size([1, 8, 77, 128]) - layer.3.k_cache: torch.Size([1, 8, 77, 128]) - layer.3.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.k_cache: torch.Size([1, 8, 77, 128]) - layer.4.v_cache: torch.Size([1, 8, 77, 128]) - layer.4.output: torch.Size([1, 77, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02707502 9.98763761 - layer.0.v_cache 0.00000027 0.00024141 - layer.1.k_cache 0.00339772 0.93173961 - layer.1.v_cache 0.00000076 0.00083080 - layer.2.k_cache 0.00114556 0.45587455 - layer.2.v_cache 0.00000112 0.00120013 - layer.3.k_cache 0.00146787 0.50737465 - layer.3.v_cache 0.00000204 0.00190570 - layer.4.k_cache 0.00318992 0.98433428 - layer.4.v_cache 0.00000300 0.00332116 - layer.4.output 0.00024419 0.11049824 - ------------------------------------------------------------------------------------- - TOTAL 0.00266143 0.95117520 - (elements=1,103,872) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1103872 -Total Bytes 239832 -BPFP 1.7381 bits/point -EBPFP 3.4762 equivalent bits/point -MSE 0.951175 ----------------------- -------------------------------------------------------- -Time: 0.491s Load: 0.005s, Pack+Encode: 0.202s, Decode+Unpack: 0.284s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 77, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 77, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9512 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample478-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample478-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample48-layer4-item1.zst (63/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample48-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 112, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 112, 128) -Output shape: (1, 112, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.output: torch.Size([1, 112, 4096]) -> torch.Size([1, 1, 112, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,476B, BPFP=0.7307 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,084B, BPFP=1.9590 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,636B, BPFP=1.3697 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,020B, BPFP=2.0940 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,164B, BPFP=1.5460 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,500B, BPFP=2.1275 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,996B, BPFP=1.6041 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 29,968B, BPFP=2.0904 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,044B, BPFP=1.3982 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,708B, BPFP=2.1420 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 85,092B, BPFP=1.4839 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.output: torch.Size([1, 112, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.288s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.output: torch.Size([1, 112, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02677278 8.05108207 - layer.0.v_cache 0.00000026 0.00023284 - layer.1.k_cache 0.00327242 0.78186492 - layer.1.v_cache 0.00000072 0.00078819 - layer.2.k_cache 0.00117792 0.42336164 - layer.2.v_cache 0.00000108 0.00111411 - layer.3.k_cache 0.00137547 0.48332480 - layer.3.v_cache 0.00000205 0.00182254 - layer.4.k_cache 0.00331726 0.91688040 - layer.4.v_cache 0.00000297 0.00305372 - layer.4.output 0.00024922 0.08160644 - ------------------------------------------------------------------------------------- - TOTAL 0.00263713 0.78499650 - (elements=1,605,632) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1605632 -Total Bytes 329688 -BPFP 1.6427 bits/point -EBPFP 3.2853 equivalent bits/point -MSE 0.784996 ----------------------- -------------------------------------------------------- -Time: 0.498s Load: 0.008s, Pack+Encode: 0.203s, Decode+Unpack: 0.288s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 112, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7850 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample48-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample48-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample485-layer4-item1.zst (64/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample485-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 78, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 78, 128) -Output shape: (1, 78, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.0.v_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.1.k_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.1.v_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.2.k_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.2.v_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.3.k_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.3.v_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.4.k_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.4.v_cache: torch.Size([1, 8, 78, 128]) -> torch.Size([1, 1, 78, 1024]) - layer.4.output: torch.Size([1, 78, 4096]) -> torch.Size([1, 1, 78, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,000B, BPFP=0.8013 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,464B, BPFP=2.0497 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,464B, BPFP=1.4487 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,388B, BPFP=2.2424 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 16,528B, BPFP=1.6554 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,236B, BPFP=2.3273 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,060B, BPFP=1.7087 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 22,656B, BPFP=2.2692 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,448B, BPFP=1.4471 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 23,372B, BPFP=2.3409 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 65,300B, BPFP=1.6351 -⌛️ [2/4] FRONTEND: Frontend time: 0.202s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 78, 128]) - layer.0.v_cache: torch.Size([1, 8, 78, 128]) - layer.1.k_cache: torch.Size([1, 8, 78, 128]) - layer.1.v_cache: torch.Size([1, 8, 78, 128]) - layer.2.k_cache: torch.Size([1, 8, 78, 128]) - layer.2.v_cache: torch.Size([1, 8, 78, 128]) - layer.3.k_cache: torch.Size([1, 8, 78, 128]) - layer.3.v_cache: torch.Size([1, 8, 78, 128]) - layer.4.k_cache: torch.Size([1, 8, 78, 128]) - layer.4.v_cache: torch.Size([1, 8, 78, 128]) - layer.4.output: torch.Size([1, 78, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 78, 128]) - layer.0.v_cache: torch.Size([1, 8, 78, 128]) - layer.1.k_cache: torch.Size([1, 8, 78, 128]) - layer.1.v_cache: torch.Size([1, 8, 78, 128]) - layer.2.k_cache: torch.Size([1, 8, 78, 128]) - layer.2.v_cache: torch.Size([1, 8, 78, 128]) - layer.3.k_cache: torch.Size([1, 8, 78, 128]) - layer.3.v_cache: torch.Size([1, 8, 78, 128]) - layer.4.k_cache: torch.Size([1, 8, 78, 128]) - layer.4.v_cache: torch.Size([1, 8, 78, 128]) - layer.4.output: torch.Size([1, 78, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02543071 8.86848645 - layer.0.v_cache 0.00000030 0.00022809 - layer.1.k_cache 0.00350441 0.93202738 - layer.1.v_cache 0.00000082 0.00083922 - layer.2.k_cache 0.00114279 0.43397556 - layer.2.v_cache 0.00000132 0.00123996 - layer.3.k_cache 0.00135780 0.48804420 - layer.3.v_cache 0.00000231 0.00203132 - layer.4.k_cache 0.00328533 0.90959715 - layer.4.v_cache 0.00000329 0.00339877 - layer.4.output 0.00022272 0.09420831 - ------------------------------------------------------------------------------------- - TOTAL 0.00254428 0.85833581 - (elements=1,118,208) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1118208 -Total Bytes 247916 -BPFP 1.7737 bits/point -EBPFP 3.5473 equivalent bits/point -MSE 0.858336 ----------------------- -------------------------------------------------------- -Time: 0.491s Load: 0.005s, Pack+Encode: 0.202s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 78, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 78, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8583 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample485-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample485-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample487-layer4-item1.zst (65/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample487-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 79, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 79, 128) -Output shape: (1, 79, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.0.v_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.1.k_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.1.v_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.2.k_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.2.v_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.3.k_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.3.v_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.4.k_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.4.v_cache: torch.Size([1, 8, 79, 128]) -> torch.Size([1, 1, 79, 1024]) - layer.4.output: torch.Size([1, 79, 4096]) -> torch.Size([1, 1, 79, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,836B, BPFP=0.7749 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 20,408B, BPFP=2.0182 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,656B, BPFP=1.4494 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,344B, BPFP=2.2097 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 16,492B, BPFP=1.6309 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 23,200B, BPFP=2.2943 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,076B, BPFP=1.6887 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 22,860B, BPFP=2.2607 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 14,532B, BPFP=1.4371 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 23,244B, BPFP=2.2987 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 62,744B, BPFP=1.5512 -⌛️ [2/4] FRONTEND: Frontend time: 0.201s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 79, 128]) - layer.0.v_cache: torch.Size([1, 8, 79, 128]) - layer.1.k_cache: torch.Size([1, 8, 79, 128]) - layer.1.v_cache: torch.Size([1, 8, 79, 128]) - layer.2.k_cache: torch.Size([1, 8, 79, 128]) - layer.2.v_cache: torch.Size([1, 8, 79, 128]) - layer.3.k_cache: torch.Size([1, 8, 79, 128]) - layer.3.v_cache: torch.Size([1, 8, 79, 128]) - layer.4.k_cache: torch.Size([1, 8, 79, 128]) - layer.4.v_cache: torch.Size([1, 8, 79, 128]) - layer.4.output: torch.Size([1, 79, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 79, 128]) - layer.0.v_cache: torch.Size([1, 8, 79, 128]) - layer.1.k_cache: torch.Size([1, 8, 79, 128]) - layer.1.v_cache: torch.Size([1, 8, 79, 128]) - layer.2.k_cache: torch.Size([1, 8, 79, 128]) - layer.2.v_cache: torch.Size([1, 8, 79, 128]) - layer.3.k_cache: torch.Size([1, 8, 79, 128]) - layer.3.v_cache: torch.Size([1, 8, 79, 128]) - layer.4.k_cache: torch.Size([1, 8, 79, 128]) - layer.4.v_cache: torch.Size([1, 8, 79, 128]) - layer.4.output: torch.Size([1, 79, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02536859 9.32836219 - layer.0.v_cache 0.00000029 0.00023291 - layer.1.k_cache 0.00331158 0.90316473 - layer.1.v_cache 0.00000077 0.00081457 - layer.2.k_cache 0.00112976 0.40498328 - layer.2.v_cache 0.00000111 0.00110454 - layer.3.k_cache 0.00133964 0.48056813 - layer.3.v_cache 0.00000223 0.00184716 - layer.4.k_cache 0.00332594 0.90493610 - layer.4.v_cache 0.00000299 0.00295881 - layer.4.output 0.00020400 0.07668245 - ------------------------------------------------------------------------------------- - TOTAL 0.00252135 0.88112159 - (elements=1,132,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1132544 -Total Bytes 245392 -BPFP 1.7334 bits/point -EBPFP 3.4668 equivalent bits/point -MSE 0.881122 ----------------------- -------------------------------------------------------- -Time: 0.492s Load: 0.005s, Pack+Encode: 0.201s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 79, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 79, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8811 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample487-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample487-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample495-layer4-item1.zst (66/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample495-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 81, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 81, 128) -Output shape: (1, 81, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) -> torch.Size([1, 1, 81, 1024]) - layer.4.output: torch.Size([1, 81, 4096]) -> torch.Size([1, 1, 81, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,508B, BPFP=0.8206 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,200B, BPFP=2.0448 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 14,956B, BPFP=1.4425 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 22,968B, BPFP=2.2153 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,044B, BPFP=1.6439 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,036B, BPFP=2.3183 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,608B, BPFP=1.6983 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,760B, BPFP=2.2917 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,072B, BPFP=1.4537 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 24,080B, BPFP=2.3225 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 68,340B, BPFP=1.6479 -⌛️ [2/4] FRONTEND: Frontend time: 0.210s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.output: torch.Size([1, 81, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.294s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 81, 128]) - layer.0.v_cache: torch.Size([1, 8, 81, 128]) - layer.1.k_cache: torch.Size([1, 8, 81, 128]) - layer.1.v_cache: torch.Size([1, 8, 81, 128]) - layer.2.k_cache: torch.Size([1, 8, 81, 128]) - layer.2.v_cache: torch.Size([1, 8, 81, 128]) - layer.3.k_cache: torch.Size([1, 8, 81, 128]) - layer.3.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.k_cache: torch.Size([1, 8, 81, 128]) - layer.4.v_cache: torch.Size([1, 8, 81, 128]) - layer.4.output: torch.Size([1, 81, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02658518 9.97926538 - layer.0.v_cache 0.00000029 0.00024410 - layer.1.k_cache 0.00348817 0.91949124 - layer.1.v_cache 0.00000087 0.00084694 - layer.2.k_cache 0.00114702 0.43464119 - layer.2.v_cache 0.00000128 0.00130890 - layer.3.k_cache 0.00135614 0.48740735 - layer.3.v_cache 0.00000252 0.00200182 - layer.4.k_cache 0.00328397 0.90464689 - layer.4.v_cache 0.00000321 0.00333593 - layer.4.output 0.00022006 0.09978286 - ------------------------------------------------------------------------------------- - TOTAL 0.00262492 0.93802294 - (elements=1,161,216) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1161216 -Total Bytes 257572 -BPFP 1.7745 bits/point -EBPFP 3.5490 equivalent bits/point -MSE 0.938023 ----------------------- -------------------------------------------------------- -Time: 0.510s Load: 0.006s, Pack+Encode: 0.210s, Decode+Unpack: 0.294s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 81, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 81, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9380 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample495-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample495-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample50-layer4-item1.zst (67/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample50-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 117, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 117, 128) -Output shape: (1, 117, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.0.v_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.1.k_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.1.v_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.2.k_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.2.v_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.3.k_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.3.v_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.4.k_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.4.v_cache: torch.Size([1, 8, 117, 128]) -> torch.Size([1, 1, 117, 1024]) - layer.4.output: torch.Size([1, 117, 4096]) -> torch.Size([1, 1, 117, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,180B, BPFP=0.7465 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,972B, BPFP=1.9346 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 20,476B, BPFP=1.3673 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,088B, BPFP=2.0759 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,948B, BPFP=1.5323 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,404B, BPFP=2.0970 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,724B, BPFP=1.5841 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,904B, BPFP=2.0636 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,784B, BPFP=1.3878 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,540B, BPFP=2.1060 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 91,244B, BPFP=1.5232 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 117, 128]) - layer.0.v_cache: torch.Size([1, 8, 117, 128]) - layer.1.k_cache: torch.Size([1, 8, 117, 128]) - layer.1.v_cache: torch.Size([1, 8, 117, 128]) - layer.2.k_cache: torch.Size([1, 8, 117, 128]) - layer.2.v_cache: torch.Size([1, 8, 117, 128]) - layer.3.k_cache: torch.Size([1, 8, 117, 128]) - layer.3.v_cache: torch.Size([1, 8, 117, 128]) - layer.4.k_cache: torch.Size([1, 8, 117, 128]) - layer.4.v_cache: torch.Size([1, 8, 117, 128]) - layer.4.output: torch.Size([1, 117, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.299s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 117, 128]) - layer.0.v_cache: torch.Size([1, 8, 117, 128]) - layer.1.k_cache: torch.Size([1, 8, 117, 128]) - layer.1.v_cache: torch.Size([1, 8, 117, 128]) - layer.2.k_cache: torch.Size([1, 8, 117, 128]) - layer.2.v_cache: torch.Size([1, 8, 117, 128]) - layer.3.k_cache: torch.Size([1, 8, 117, 128]) - layer.3.v_cache: torch.Size([1, 8, 117, 128]) - layer.4.k_cache: torch.Size([1, 8, 117, 128]) - layer.4.v_cache: torch.Size([1, 8, 117, 128]) - layer.4.output: torch.Size([1, 117, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02657393 8.34990360 - layer.0.v_cache 0.00000026 0.00022092 - layer.1.k_cache 0.00326448 0.79308345 - layer.1.v_cache 0.00000079 0.00081481 - layer.2.k_cache 0.00117087 0.41868021 - layer.2.v_cache 0.00000122 0.00119063 - layer.3.k_cache 0.00141366 0.47726753 - layer.3.v_cache 0.00000241 0.00189131 - layer.4.k_cache 0.00339367 0.89367193 - layer.4.v_cache 0.00000311 0.00309886 - layer.4.output 0.00025967 0.09219210 - ------------------------------------------------------------------------------------- - TOTAL 0.00263308 0.80775655 - (elements=1,677,312) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1677312 -Total Bytes 344264 -BPFP 1.6420 bits/point -EBPFP 3.2840 equivalent bits/point -MSE 0.807757 ----------------------- -------------------------------------------------------- -Time: 0.516s Load: 0.007s, Pack+Encode: 0.209s, Decode+Unpack: 0.299s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 117, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 117, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8078 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample50-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample50-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample516-layer4-item1.zst (68/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample516-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 84, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 84, 128) -Output shape: (1, 84, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.0.v_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.1.k_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.1.v_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.2.k_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.2.v_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.3.k_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.3.v_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.4.k_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.4.v_cache: torch.Size([1, 8, 84, 128]) -> torch.Size([1, 1, 84, 1024]) - layer.4.output: torch.Size([1, 84, 4096]) -> torch.Size([1, 1, 84, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 8,152B, BPFP=0.7582 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 21,320B, BPFP=1.9829 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 15,124B, BPFP=1.4066 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 23,616B, BPFP=2.1964 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 17,364B, BPFP=1.6150 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 24,288B, BPFP=2.2589 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 17,896B, BPFP=1.6644 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 23,992B, BPFP=2.2314 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 15,332B, BPFP=1.4260 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 24,624B, BPFP=2.2902 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 67,964B, BPFP=1.5803 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 84, 128]) - layer.0.v_cache: torch.Size([1, 8, 84, 128]) - layer.1.k_cache: torch.Size([1, 8, 84, 128]) - layer.1.v_cache: torch.Size([1, 8, 84, 128]) - layer.2.k_cache: torch.Size([1, 8, 84, 128]) - layer.2.v_cache: torch.Size([1, 8, 84, 128]) - layer.3.k_cache: torch.Size([1, 8, 84, 128]) - layer.3.v_cache: torch.Size([1, 8, 84, 128]) - layer.4.k_cache: torch.Size([1, 8, 84, 128]) - layer.4.v_cache: torch.Size([1, 8, 84, 128]) - layer.4.output: torch.Size([1, 84, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.295s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 84, 128]) - layer.0.v_cache: torch.Size([1, 8, 84, 128]) - layer.1.k_cache: torch.Size([1, 8, 84, 128]) - layer.1.v_cache: torch.Size([1, 8, 84, 128]) - layer.2.k_cache: torch.Size([1, 8, 84, 128]) - layer.2.v_cache: torch.Size([1, 8, 84, 128]) - layer.3.k_cache: torch.Size([1, 8, 84, 128]) - layer.3.v_cache: torch.Size([1, 8, 84, 128]) - layer.4.k_cache: torch.Size([1, 8, 84, 128]) - layer.4.v_cache: torch.Size([1, 8, 84, 128]) - layer.4.output: torch.Size([1, 84, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02585370 9.94118972 - layer.0.v_cache 0.00000028 0.00023349 - layer.1.k_cache 0.00342466 0.91111465 - layer.1.v_cache 0.00000080 0.00081131 - layer.2.k_cache 0.00116354 0.40288008 - layer.2.v_cache 0.00000112 0.00111655 - layer.3.k_cache 0.00133515 0.46593003 - layer.3.v_cache 0.00000214 0.00181124 - layer.4.k_cache 0.00358854 0.85730734 - layer.4.v_cache 0.00000313 0.00302089 - layer.4.output 0.00016584 0.07510928 - ------------------------------------------------------------------------------------- - TOTAL 0.00257403 0.92041803 - (elements=1,204,224) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1204224 -Total Bytes 259672 -BPFP 1.7251 bits/point -EBPFP 3.4501 equivalent bits/point -MSE 0.920418 ----------------------- -------------------------------------------------------- -Time: 0.508s Load: 0.005s, Pack+Encode: 0.207s, Decode+Unpack: 0.295s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 84, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 84, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9204 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample516-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample516-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample52-layer4-item1.zst (69/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample52-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 133, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 133, 128) -Output shape: (1, 133, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.0.v_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.1.k_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.1.v_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.2.k_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.2.v_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.3.k_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.3.v_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.4.k_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.4.v_cache: torch.Size([1, 8, 133, 128]) -> torch.Size([1, 1, 133, 1024]) - layer.4.output: torch.Size([1, 133, 4096]) -> torch.Size([1, 1, 133, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 13,244B, BPFP=0.7780 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 34,276B, BPFP=2.0134 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 24,184B, BPFP=1.4206 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 37,032B, BPFP=2.1753 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 27,104B, BPFP=1.5921 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 37,720B, BPFP=2.2157 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 28,348B, BPFP=1.6652 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 37,144B, BPFP=2.1819 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 24,512B, BPFP=1.4398 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 37,888B, BPFP=2.2256 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 106,596B, BPFP=1.5654 -⌛️ [2/4] FRONTEND: Frontend time: 0.254s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 133, 128]) - layer.0.v_cache: torch.Size([1, 8, 133, 128]) - layer.1.k_cache: torch.Size([1, 8, 133, 128]) - layer.1.v_cache: torch.Size([1, 8, 133, 128]) - layer.2.k_cache: torch.Size([1, 8, 133, 128]) - layer.2.v_cache: torch.Size([1, 8, 133, 128]) - layer.3.k_cache: torch.Size([1, 8, 133, 128]) - layer.3.v_cache: torch.Size([1, 8, 133, 128]) - layer.4.k_cache: torch.Size([1, 8, 133, 128]) - layer.4.v_cache: torch.Size([1, 8, 133, 128]) - layer.4.output: torch.Size([1, 133, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.394s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 133, 128]) - layer.0.v_cache: torch.Size([1, 8, 133, 128]) - layer.1.k_cache: torch.Size([1, 8, 133, 128]) - layer.1.v_cache: torch.Size([1, 8, 133, 128]) - layer.2.k_cache: torch.Size([1, 8, 133, 128]) - layer.2.v_cache: torch.Size([1, 8, 133, 128]) - layer.3.k_cache: torch.Size([1, 8, 133, 128]) - layer.3.v_cache: torch.Size([1, 8, 133, 128]) - layer.4.k_cache: torch.Size([1, 8, 133, 128]) - layer.4.v_cache: torch.Size([1, 8, 133, 128]) - layer.4.output: torch.Size([1, 133, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02254625 9.27306578 - layer.0.v_cache 0.00000028 0.00024313 - layer.1.k_cache 0.00324298 0.82082086 - layer.1.v_cache 0.00000087 0.00077756 - layer.2.k_cache 0.00114197 0.39829865 - layer.2.v_cache 0.00000110 0.00115262 - layer.3.k_cache 0.00134409 0.45572003 - layer.3.v_cache 0.00000211 0.00183989 - layer.4.k_cache 0.00375619 0.87933051 - layer.4.v_cache 0.00000306 0.00316477 - layer.4.output 0.00014515 0.06818644 - ------------------------------------------------------------------------------------- - TOTAL 0.00232996 0.86479711 - (elements=1,906,688) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1906688 -Total Bytes 408048 -BPFP 1.7121 bits/point -EBPFP 3.4241 equivalent bits/point -MSE 0.864797 ----------------------- -------------------------------------------------------- -Time: 0.656s Load: 0.007s, Pack+Encode: 0.254s, Decode+Unpack: 0.394s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 133, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 133, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8648 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample52-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample52-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample53-layer4-item1.zst (70/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample53-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 125, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 125, 128) -Output shape: (1, 125, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.0.v_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.1.k_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.1.v_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.2.k_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.2.v_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.3.k_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.3.v_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.4.k_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.4.v_cache: torch.Size([1, 8, 125, 128]) -> torch.Size([1, 1, 125, 1024]) - layer.4.output: torch.Size([1, 125, 4096]) -> torch.Size([1, 1, 125, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,748B, BPFP=0.7342 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,948B, BPFP=1.8093 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 21,552B, BPFP=1.3470 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,792B, BPFP=1.9245 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,892B, BPFP=1.4932 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,316B, BPFP=1.9572 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,544B, BPFP=1.5340 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,956B, BPFP=1.9347 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 21,924B, BPFP=1.3702 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,492B, BPFP=1.9683 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 93,260B, BPFP=1.4572 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 125, 128]) - layer.0.v_cache: torch.Size([1, 8, 125, 128]) - layer.1.k_cache: torch.Size([1, 8, 125, 128]) - layer.1.v_cache: torch.Size([1, 8, 125, 128]) - layer.2.k_cache: torch.Size([1, 8, 125, 128]) - layer.2.v_cache: torch.Size([1, 8, 125, 128]) - layer.3.k_cache: torch.Size([1, 8, 125, 128]) - layer.3.v_cache: torch.Size([1, 8, 125, 128]) - layer.4.k_cache: torch.Size([1, 8, 125, 128]) - layer.4.v_cache: torch.Size([1, 8, 125, 128]) - layer.4.output: torch.Size([1, 125, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.299s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 125, 128]) - layer.0.v_cache: torch.Size([1, 8, 125, 128]) - layer.1.k_cache: torch.Size([1, 8, 125, 128]) - layer.1.v_cache: torch.Size([1, 8, 125, 128]) - layer.2.k_cache: torch.Size([1, 8, 125, 128]) - layer.2.v_cache: torch.Size([1, 8, 125, 128]) - layer.3.k_cache: torch.Size([1, 8, 125, 128]) - layer.3.v_cache: torch.Size([1, 8, 125, 128]) - layer.4.k_cache: torch.Size([1, 8, 125, 128]) - layer.4.v_cache: torch.Size([1, 8, 125, 128]) - layer.4.output: torch.Size([1, 125, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02456201 7.61295117 - layer.0.v_cache 0.00000027 0.00022541 - layer.1.k_cache 0.00318185 0.68950616 - layer.1.v_cache 0.00000072 0.00077233 - layer.2.k_cache 0.00113729 0.37776004 - layer.2.v_cache 0.00000107 0.00108339 - layer.3.k_cache 0.00136705 0.44739850 - layer.3.v_cache 0.00000203 0.00180767 - layer.4.k_cache 0.00364986 0.82587323 - layer.4.v_cache 0.00000297 0.00291252 - layer.4.output 0.00019454 0.06815363 - ------------------------------------------------------------------------------------- - TOTAL 0.00247738 0.73092178 - (elements=1,792,000) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1792000 -Total Bytes 350424 -BPFP 1.5644 bits/point -EBPFP 3.1288 equivalent bits/point -MSE 0.730922 ----------------------- -------------------------------------------------------- -Time: 0.517s Load: 0.008s, Pack+Encode: 0.209s, Decode+Unpack: 0.299s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 125, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 125, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7309 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample53-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample53-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample543-layer4-item1.zst (71/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample543-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 71, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 71, 128) -Output shape: (1, 71, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.output: torch.Size([1, 71, 4096]) -> torch.Size([1, 1, 71, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,340B, BPFP=0.8077 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,352B, BPFP=2.0194 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,316B, BPFP=1.4652 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,296B, BPFP=2.2333 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,724B, BPFP=1.6202 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,896B, BPFP=2.2993 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,364B, BPFP=1.6906 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,448B, BPFP=2.2500 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,124B, BPFP=1.4441 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,856B, BPFP=2.2949 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 58,932B, BPFP=1.6211 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.output: torch.Size([1, 71, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.output: torch.Size([1, 71, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02849298 11.67201598 - layer.0.v_cache 0.00000028 0.00024363 - layer.1.k_cache 0.00376352 0.93433692 - layer.1.v_cache 0.00000080 0.00087206 - layer.2.k_cache 0.00117123 0.42232484 - layer.2.v_cache 0.00000118 0.00115446 - layer.3.k_cache 0.00138233 0.49834947 - layer.3.v_cache 0.00000223 0.00195222 - layer.4.k_cache 0.00337686 0.94572782 - layer.4.v_cache 0.00000300 0.00317550 - layer.4.output 0.00019204 0.09690620 - ------------------------------------------------------------------------------------- - TOTAL 0.00278304 1.06198412 - (elements=1,017,856) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1017856 -Total Bytes 223648 -BPFP 1.7578 bits/point -EBPFP 3.5156 equivalent bits/point -MSE 1.061984 ----------------------- -------------------------------------------------------- -Time: 0.510s Load: 0.005s, Pack+Encode: 0.209s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 71, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.0620 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample543-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample543-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample561-layer4-item1.zst (72/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample561-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 68, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 68, 128) -Output shape: (1, 68, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.output: torch.Size([1, 68, 4096]) -> torch.Size([1, 1, 68, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,456B, BPFP=0.8566 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 17,624B, BPFP=2.0248 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 12,468B, BPFP=1.4324 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 19,932B, BPFP=2.2900 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,396B, BPFP=1.6540 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,400B, BPFP=2.3438 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 14,684B, BPFP=1.6870 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 19,976B, BPFP=2.2950 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 12,604B, BPFP=1.4481 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,568B, BPFP=2.3631 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 55,668B, BPFP=1.5989 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.output: torch.Size([1, 68, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.295s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.output: torch.Size([1, 68, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02799314 10.57614764 - layer.0.v_cache 0.00000027 0.00024948 - layer.1.k_cache 0.00364027 1.09382102 - layer.1.v_cache 0.00000080 0.00094538 - layer.2.k_cache 0.00115812 0.45566794 - layer.2.v_cache 0.00000117 0.00128385 - layer.3.k_cache 0.00137295 0.52245723 - layer.3.v_cache 0.00000227 0.00215734 - layer.4.k_cache 0.00327663 0.99662803 - layer.4.v_cache 0.00000341 0.00368257 - layer.4.output 0.00018782 0.09915378 - ------------------------------------------------------------------------------------- - TOTAL 0.00272859 1.00354683 - (elements=974,848) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 974848 -Total Bytes 215776 -BPFP 1.7707 bits/point -EBPFP 3.5415 equivalent bits/point -MSE 1.003547 ----------------------- -------------------------------------------------------- -Time: 0.506s Load: 0.005s, Pack+Encode: 0.207s, Decode+Unpack: 0.295s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 68, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.0035 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample561-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample561-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample570-layer4-item1.zst (73/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample570-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 72, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 72, 128) -Output shape: (1, 72, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.output: torch.Size([1, 72, 4096]) -> torch.Size([1, 1, 72, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,416B, BPFP=0.8047 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,628B, BPFP=2.0213 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,360B, BPFP=1.4497 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,360B, BPFP=2.2092 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,056B, BPFP=1.6337 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,024B, BPFP=2.2812 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,700B, BPFP=1.7036 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,556B, BPFP=2.2305 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,248B, BPFP=1.4375 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,720B, BPFP=2.2483 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 57,628B, BPFP=1.5633 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.output: torch.Size([1, 72, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.284s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.output: torch.Size([1, 72, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02695644 11.22194163 - layer.0.v_cache 0.00000028 0.00024297 - layer.1.k_cache 0.00362044 0.95270824 - layer.1.v_cache 0.00000088 0.00085609 - layer.2.k_cache 0.00116887 0.43695076 - layer.2.v_cache 0.00000122 0.00120906 - layer.3.k_cache 0.00138034 0.49947453 - layer.3.v_cache 0.00000238 0.00198827 - layer.4.k_cache 0.00335768 0.98594475 - layer.4.v_cache 0.00000294 0.00309665 - layer.4.output 0.00021991 0.08096606 - ------------------------------------------------------------------------------------- - TOTAL 0.00266936 1.03059123 - (elements=1,032,192) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1032192 -Total Bytes 223696 -BPFP 1.7338 bits/point -EBPFP 3.4675 equivalent bits/point -MSE 1.030591 ----------------------- -------------------------------------------------------- -Time: 0.492s Load: 0.005s, Pack+Encode: 0.203s, Decode+Unpack: 0.284s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 72, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.0306 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample570-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample570-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample581-layer4-item1.zst (74/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample581-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 73, 128) -Output shape: (1, 73, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.output: torch.Size([1, 73, 4096]) -> torch.Size([1, 1, 73, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,384B, BPFP=0.7902 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,808B, BPFP=2.0128 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,380B, BPFP=1.4319 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,416B, BPFP=2.1849 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,920B, BPFP=1.5967 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,004B, BPFP=2.2479 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,516B, BPFP=1.6605 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,756B, BPFP=2.2213 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,244B, BPFP=1.4174 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 21,248B, BPFP=2.2740 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 57,884B, BPFP=1.5487 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02610407 10.07462008 - layer.0.v_cache 0.00000028 0.00024220 - layer.1.k_cache 0.00357364 0.91365020 - layer.1.v_cache 0.00000075 0.00082768 - layer.2.k_cache 0.00114681 0.42975232 - layer.2.v_cache 0.00000108 0.00112744 - layer.3.k_cache 0.00136272 0.48315848 - layer.3.v_cache 0.00000212 0.00181772 - layer.4.k_cache 0.00341161 0.95716461 - layer.4.v_cache 0.00000293 0.00307329 - layer.4.output 0.00018354 0.07893837 - ------------------------------------------------------------------------------------- - TOTAL 0.00259573 0.94151339 - (elements=1,046,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1046528 -Total Bytes 224560 -BPFP 1.7166 bits/point -EBPFP 3.4332 equivalent bits/point -MSE 0.941513 ----------------------- -------------------------------------------------------- -Time: 0.494s Load: 0.005s, Pack+Encode: 0.205s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9415 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample581-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample581-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample584-layer4-item1.zst (75/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample584-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 72, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 72, 128) -Output shape: (1, 72, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) -> torch.Size([1, 1, 72, 1024]) - layer.4.output: torch.Size([1, 72, 4096]) -> torch.Size([1, 1, 72, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,148B, BPFP=0.7756 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,376B, BPFP=1.9939 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,288B, BPFP=1.4418 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 19,988B, BPFP=2.1688 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,904B, BPFP=1.6172 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,648B, BPFP=2.2405 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,476B, BPFP=1.6793 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,296B, BPFP=2.2023 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,140B, BPFP=1.4258 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,772B, BPFP=2.2539 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 57,164B, BPFP=1.5507 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.output: torch.Size([1, 72, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.284s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 72, 128]) - layer.0.v_cache: torch.Size([1, 8, 72, 128]) - layer.1.k_cache: torch.Size([1, 8, 72, 128]) - layer.1.v_cache: torch.Size([1, 8, 72, 128]) - layer.2.k_cache: torch.Size([1, 8, 72, 128]) - layer.2.v_cache: torch.Size([1, 8, 72, 128]) - layer.3.k_cache: torch.Size([1, 8, 72, 128]) - layer.3.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.k_cache: torch.Size([1, 8, 72, 128]) - layer.4.v_cache: torch.Size([1, 8, 72, 128]) - layer.4.output: torch.Size([1, 72, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02727311 10.37810686 - layer.0.v_cache 0.00000029 0.00024926 - layer.1.k_cache 0.00335621 0.92697101 - layer.1.v_cache 0.00000072 0.00077559 - layer.2.k_cache 0.00111533 0.41395198 - layer.2.v_cache 0.00000112 0.00108389 - layer.3.k_cache 0.00138591 0.48073859 - layer.3.v_cache 0.00000206 0.00175002 - layer.4.k_cache 0.00323834 0.90059302 - layer.4.v_cache 0.00000293 0.00298952 - layer.4.output 0.00022510 0.08378300 - ------------------------------------------------------------------------------------- - TOTAL 0.00266260 0.96016727 - (elements=1,032,192) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1032192 -Total Bytes 221200 -BPFP 1.7144 bits/point -EBPFP 3.4288 equivalent bits/point -MSE 0.960167 ----------------------- -------------------------------------------------------- -Time: 0.494s Load: 0.006s, Pack+Encode: 0.204s, Decode+Unpack: 0.284s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 72, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 72, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9602 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample584-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample584-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample60-layer4-item1.zst (76/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample60-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 115, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 115, 128) -Output shape: (1, 115, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.0.v_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.1.k_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.1.v_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.2.k_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.2.v_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.3.k_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.3.v_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.4.k_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.4.v_cache: torch.Size([1, 8, 115, 128]) -> torch.Size([1, 1, 115, 1024]) - layer.4.output: torch.Size([1, 115, 4096]) -> torch.Size([1, 1, 115, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,124B, BPFP=0.7557 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,816B, BPFP=1.9576 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 20,068B, BPFP=1.3633 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,912B, BPFP=2.1000 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,592B, BPFP=1.5348 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,128B, BPFP=2.1147 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,304B, BPFP=1.5832 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,596B, BPFP=2.0785 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,480B, BPFP=1.3913 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,416B, BPFP=2.1342 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 91,424B, BPFP=1.5527 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 115, 128]) - layer.0.v_cache: torch.Size([1, 8, 115, 128]) - layer.1.k_cache: torch.Size([1, 8, 115, 128]) - layer.1.v_cache: torch.Size([1, 8, 115, 128]) - layer.2.k_cache: torch.Size([1, 8, 115, 128]) - layer.2.v_cache: torch.Size([1, 8, 115, 128]) - layer.3.k_cache: torch.Size([1, 8, 115, 128]) - layer.3.v_cache: torch.Size([1, 8, 115, 128]) - layer.4.k_cache: torch.Size([1, 8, 115, 128]) - layer.4.v_cache: torch.Size([1, 8, 115, 128]) - layer.4.output: torch.Size([1, 115, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.290s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 115, 128]) - layer.0.v_cache: torch.Size([1, 8, 115, 128]) - layer.1.k_cache: torch.Size([1, 8, 115, 128]) - layer.1.v_cache: torch.Size([1, 8, 115, 128]) - layer.2.k_cache: torch.Size([1, 8, 115, 128]) - layer.2.v_cache: torch.Size([1, 8, 115, 128]) - layer.3.k_cache: torch.Size([1, 8, 115, 128]) - layer.3.v_cache: torch.Size([1, 8, 115, 128]) - layer.4.k_cache: torch.Size([1, 8, 115, 128]) - layer.4.v_cache: torch.Size([1, 8, 115, 128]) - layer.4.output: torch.Size([1, 115, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02486876 8.97977136 - layer.0.v_cache 0.00000027 0.00023950 - layer.1.k_cache 0.00345256 0.76043648 - layer.1.v_cache 0.00000081 0.00087153 - layer.2.k_cache 0.00114556 0.40655750 - layer.2.v_cache 0.00000111 0.00119688 - layer.3.k_cache 0.00139466 0.47691262 - layer.3.v_cache 0.00000215 0.00198972 - layer.4.k_cache 0.00339940 0.86541781 - layer.4.v_cache 0.00000318 0.00347406 - layer.4.output 0.00018696 0.07605929 - ------------------------------------------------------------------------------------- - TOTAL 0.00250116 0.84293605 - (elements=1,648,640) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1648640 -Total Bytes 341860 -BPFP 1.6589 bits/point -EBPFP 3.3177 equivalent bits/point -MSE 0.842936 ----------------------- -------------------------------------------------------- -Time: 0.504s Load: 0.007s, Pack+Encode: 0.207s, Decode+Unpack: 0.290s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 115, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 115, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8429 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample60-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample60-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample600-layer4-item1.zst (77/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample600-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 71, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 71, 128) -Output shape: (1, 71, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) -> torch.Size([1, 1, 71, 1024]) - layer.4.output: torch.Size([1, 71, 4096]) -> torch.Size([1, 1, 71, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,684B, BPFP=0.8455 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,624B, BPFP=2.0493 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,244B, BPFP=1.4573 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,376B, BPFP=2.2421 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,776B, BPFP=1.6259 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,800B, BPFP=2.2887 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,400B, BPFP=1.6945 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,556B, BPFP=2.2619 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,128B, BPFP=1.4445 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,704B, BPFP=2.2782 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 57,784B, BPFP=1.5896 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.output: torch.Size([1, 71, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.286s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 71, 128]) - layer.0.v_cache: torch.Size([1, 8, 71, 128]) - layer.1.k_cache: torch.Size([1, 8, 71, 128]) - layer.1.v_cache: torch.Size([1, 8, 71, 128]) - layer.2.k_cache: torch.Size([1, 8, 71, 128]) - layer.2.v_cache: torch.Size([1, 8, 71, 128]) - layer.3.k_cache: torch.Size([1, 8, 71, 128]) - layer.3.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.k_cache: torch.Size([1, 8, 71, 128]) - layer.4.v_cache: torch.Size([1, 8, 71, 128]) - layer.4.output: torch.Size([1, 71, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02663550 10.78214404 - layer.0.v_cache 0.00000027 0.00023833 - layer.1.k_cache 0.00355294 0.94462521 - layer.1.v_cache 0.00000089 0.00088557 - layer.2.k_cache 0.00117088 0.43783731 - layer.2.v_cache 0.00000116 0.00120791 - layer.3.k_cache 0.00139091 0.49381261 - layer.3.v_cache 0.00000218 0.00195015 - layer.4.k_cache 0.00325425 0.97998992 - layer.4.v_cache 0.00000306 0.00318141 - layer.4.output 0.00022008 0.09225193 - ------------------------------------------------------------------------------------- - TOTAL 0.00263517 1.00106287 - (elements=1,017,856) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1017856 -Total Bytes 223076 -BPFP 1.7533 bits/point -EBPFP 3.5066 equivalent bits/point -MSE 1.001063 ----------------------- -------------------------------------------------------- -Time: 0.498s Load: 0.005s, Pack+Encode: 0.207s, Decode+Unpack: 0.286s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 71, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 71, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 1.0011 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample600-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample600-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample622-layer4-item1.zst (78/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample622-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 73, 128) -Output shape: (1, 73, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.output: torch.Size([1, 73, 4096]) -> torch.Size([1, 1, 73, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,512B, BPFP=0.8039 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,856B, BPFP=2.0180 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,576B, BPFP=1.4529 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,636B, BPFP=2.2085 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,996B, BPFP=1.6049 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,284B, BPFP=2.2778 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,636B, BPFP=1.6734 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,852B, BPFP=2.2316 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,328B, BPFP=1.4264 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 21,204B, BPFP=2.2693 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 59,072B, BPFP=1.5805 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02450750 10.11157394 - layer.0.v_cache 0.00000028 0.00025238 - layer.1.k_cache 0.00347943 0.97154257 - layer.1.v_cache 0.00000079 0.00086689 - layer.2.k_cache 0.00114165 0.44193158 - layer.2.v_cache 0.00000118 0.00122659 - layer.3.k_cache 0.00136377 0.49791106 - layer.3.v_cache 0.00000212 0.00196466 - layer.4.k_cache 0.00321585 0.93832920 - layer.4.v_cache 0.00000299 0.00323175 - layer.4.output 0.00023923 0.09208919 - ------------------------------------------------------------------------------------- - TOTAL 0.00247661 0.95265624 - (elements=1,046,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1046528 -Total Bytes 226952 -BPFP 1.7349 bits/point -EBPFP 3.4698 equivalent bits/point -MSE 0.952656 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.005s, Pack+Encode: 0.205s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9527 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample622-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample622-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample65-layer4-item1.zst (79/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample65-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 112, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 112, 128) -Output shape: (1, 112, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) -> torch.Size([1, 1, 112, 1024]) - layer.4.output: torch.Size([1, 112, 4096]) -> torch.Size([1, 1, 112, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,640B, BPFP=0.8119 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,732B, BPFP=2.0042 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,808B, BPFP=1.3817 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,696B, BPFP=2.1412 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,176B, BPFP=1.5469 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,904B, BPFP=2.1557 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,936B, BPFP=1.5999 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,472B, BPFP=2.1256 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,008B, BPFP=1.3956 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,052B, BPFP=2.1660 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 91,844B, BPFP=1.6016 -⌛️ [2/4] FRONTEND: Frontend time: 0.206s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.output: torch.Size([1, 112, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.288s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 112, 128]) - layer.0.v_cache: torch.Size([1, 8, 112, 128]) - layer.1.k_cache: torch.Size([1, 8, 112, 128]) - layer.1.v_cache: torch.Size([1, 8, 112, 128]) - layer.2.k_cache: torch.Size([1, 8, 112, 128]) - layer.2.v_cache: torch.Size([1, 8, 112, 128]) - layer.3.k_cache: torch.Size([1, 8, 112, 128]) - layer.3.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.k_cache: torch.Size([1, 8, 112, 128]) - layer.4.v_cache: torch.Size([1, 8, 112, 128]) - layer.4.output: torch.Size([1, 112, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02850304 7.98957171 - layer.0.v_cache 0.00000027 0.00025270 - layer.1.k_cache 0.00329348 0.80474949 - layer.1.v_cache 0.00000088 0.00097220 - layer.2.k_cache 0.00114370 0.41144143 - layer.2.v_cache 0.00000135 0.00125134 - layer.3.k_cache 0.00133332 0.49096516 - layer.3.v_cache 0.00000257 0.00218325 - layer.4.k_cache 0.00339413 0.82364934 - layer.4.v_cache 0.00000321 0.00344092 - layer.4.output 0.00019581 0.07560024 - ------------------------------------------------------------------------------------- - TOTAL 0.00274709 0.77363418 - (elements=1,605,632) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1605632 -Total Bytes 340268 -BPFP 1.6954 bits/point -EBPFP 3.3907 equivalent bits/point -MSE 0.773634 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.007s, Pack+Encode: 0.206s, Decode+Unpack: 0.288s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 112, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 112, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7736 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample65-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample65-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample656-layer4-item1.zst (80/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample656-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 68, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 68, 128) -Output shape: (1, 68, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) -> torch.Size([1, 1, 68, 1024]) - layer.4.output: torch.Size([1, 68, 4096]) -> torch.Size([1, 1, 68, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,392B, BPFP=0.8493 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 17,700B, BPFP=2.0335 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 12,360B, BPFP=1.4200 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 19,744B, BPFP=2.2684 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,008B, BPFP=1.6094 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,224B, BPFP=2.3235 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 14,412B, BPFP=1.6558 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 19,876B, BPFP=2.2835 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 12,420B, BPFP=1.4269 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,248B, BPFP=2.3263 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 55,340B, BPFP=1.5895 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.output: torch.Size([1, 68, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.282s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 68, 128]) - layer.0.v_cache: torch.Size([1, 8, 68, 128]) - layer.1.k_cache: torch.Size([1, 8, 68, 128]) - layer.1.v_cache: torch.Size([1, 8, 68, 128]) - layer.2.k_cache: torch.Size([1, 8, 68, 128]) - layer.2.v_cache: torch.Size([1, 8, 68, 128]) - layer.3.k_cache: torch.Size([1, 8, 68, 128]) - layer.3.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.k_cache: torch.Size([1, 8, 68, 128]) - layer.4.v_cache: torch.Size([1, 8, 68, 128]) - layer.4.output: torch.Size([1, 68, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02721363 10.67794261 - layer.0.v_cache 0.00000029 0.00024124 - layer.1.k_cache 0.00369932 0.97898483 - layer.1.v_cache 0.00000081 0.00083658 - layer.2.k_cache 0.00117793 0.43637096 - layer.2.v_cache 0.00000112 0.00117191 - layer.3.k_cache 0.00140028 0.50686909 - layer.3.v_cache 0.00000224 0.00189534 - layer.4.k_cache 0.00315349 0.95664069 - layer.4.v_cache 0.00000301 0.00311514 - layer.4.output 0.00021496 0.09503825 - ------------------------------------------------------------------------------------- - TOTAL 0.00267943 0.99601581 - (elements=974,848) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 974848 -Total Bytes 213724 -BPFP 1.7539 bits/point -EBPFP 3.5078 equivalent bits/point -MSE 0.996016 ----------------------- -------------------------------------------------------- -Time: 0.491s Load: 0.005s, Pack+Encode: 0.205s, Decode+Unpack: 0.282s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 68, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 68, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9960 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample656-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample656-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample66-layer4-item1.zst (81/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample66-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 136, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.009s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 136, 128) -Output shape: (1, 136, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.0.v_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.1.k_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.1.v_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.2.k_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.2.v_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.3.k_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.3.v_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.4.k_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.4.v_cache: torch.Size([1, 8, 136, 128]) -> torch.Size([1, 1, 136, 1024]) - layer.4.output: torch.Size([1, 136, 4096]) -> torch.Size([1, 1, 136, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 14,164B, BPFP=0.8136 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 35,212B, BPFP=2.0227 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 25,008B, BPFP=1.4366 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 37,792B, BPFP=2.1710 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 28,060B, BPFP=1.6119 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 38,480B, BPFP=2.2105 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 29,336B, BPFP=1.6852 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 38,088B, BPFP=2.1880 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 25,168B, BPFP=1.4458 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 38,796B, BPFP=2.2286 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 113,948B, BPFP=1.6364 -⌛️ [2/4] FRONTEND: Frontend time: 0.260s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 136, 128]) - layer.0.v_cache: torch.Size([1, 8, 136, 128]) - layer.1.k_cache: torch.Size([1, 8, 136, 128]) - layer.1.v_cache: torch.Size([1, 8, 136, 128]) - layer.2.k_cache: torch.Size([1, 8, 136, 128]) - layer.2.v_cache: torch.Size([1, 8, 136, 128]) - layer.3.k_cache: torch.Size([1, 8, 136, 128]) - layer.3.v_cache: torch.Size([1, 8, 136, 128]) - layer.4.k_cache: torch.Size([1, 8, 136, 128]) - layer.4.v_cache: torch.Size([1, 8, 136, 128]) - layer.4.output: torch.Size([1, 136, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.388s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 136, 128]) - layer.0.v_cache: torch.Size([1, 8, 136, 128]) - layer.1.k_cache: torch.Size([1, 8, 136, 128]) - layer.1.v_cache: torch.Size([1, 8, 136, 128]) - layer.2.k_cache: torch.Size([1, 8, 136, 128]) - layer.2.v_cache: torch.Size([1, 8, 136, 128]) - layer.3.k_cache: torch.Size([1, 8, 136, 128]) - layer.3.v_cache: torch.Size([1, 8, 136, 128]) - layer.4.k_cache: torch.Size([1, 8, 136, 128]) - layer.4.v_cache: torch.Size([1, 8, 136, 128]) - layer.4.output: torch.Size([1, 136, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02725438 8.52759238 - layer.0.v_cache 0.00000027 0.00023624 - layer.1.k_cache 0.00321745 0.75290938 - layer.1.v_cache 0.00000084 0.00078604 - layer.2.k_cache 0.00114365 0.39558932 - layer.2.v_cache 0.00000128 0.00110981 - layer.3.k_cache 0.00137918 0.44579854 - layer.3.v_cache 0.00000219 0.00182889 - layer.4.k_cache 0.00340721 0.85186812 - layer.4.v_cache 0.00000338 0.00312122 - layer.4.output 0.00019585 0.07722024 - ------------------------------------------------------------------------------------- - TOTAL 0.00265666 0.80640863 - (elements=1,949,696) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1949696 -Total Bytes 424052 -BPFP 1.7400 bits/point -EBPFP 3.4799 equivalent bits/point -MSE 0.806409 ----------------------- -------------------------------------------------------- -Time: 0.657s Load: 0.009s, Pack+Encode: 0.260s, Decode+Unpack: 0.388s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 136, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 136, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8064 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample66-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample66-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample663-layer4-item1.zst (82/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample663-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 67, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.004s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 67, 128) -Output shape: (1, 67, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.0.v_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.1.k_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.1.v_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.2.k_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.2.v_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.3.k_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.3.v_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.4.k_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.4.v_cache: torch.Size([1, 8, 67, 128]) -> torch.Size([1, 1, 67, 1024]) - layer.4.output: torch.Size([1, 67, 4096]) -> torch.Size([1, 1, 67, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,120B, BPFP=0.8302 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 17,464B, BPFP=2.0364 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 12,348B, BPFP=1.4398 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 19,452B, BPFP=2.2682 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,192B, BPFP=1.6549 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,296B, BPFP=2.3666 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 14,716B, BPFP=1.7160 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 19,884B, BPFP=2.3186 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 12,352B, BPFP=1.4403 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,120B, BPFP=2.3461 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 55,224B, BPFP=1.6098 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 67, 128]) - layer.0.v_cache: torch.Size([1, 8, 67, 128]) - layer.1.k_cache: torch.Size([1, 8, 67, 128]) - layer.1.v_cache: torch.Size([1, 8, 67, 128]) - layer.2.k_cache: torch.Size([1, 8, 67, 128]) - layer.2.v_cache: torch.Size([1, 8, 67, 128]) - layer.3.k_cache: torch.Size([1, 8, 67, 128]) - layer.3.v_cache: torch.Size([1, 8, 67, 128]) - layer.4.k_cache: torch.Size([1, 8, 67, 128]) - layer.4.v_cache: torch.Size([1, 8, 67, 128]) - layer.4.output: torch.Size([1, 67, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.284s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 67, 128]) - layer.0.v_cache: torch.Size([1, 8, 67, 128]) - layer.1.k_cache: torch.Size([1, 8, 67, 128]) - layer.1.v_cache: torch.Size([1, 8, 67, 128]) - layer.2.k_cache: torch.Size([1, 8, 67, 128]) - layer.2.v_cache: torch.Size([1, 8, 67, 128]) - layer.3.k_cache: torch.Size([1, 8, 67, 128]) - layer.3.v_cache: torch.Size([1, 8, 67, 128]) - layer.4.k_cache: torch.Size([1, 8, 67, 128]) - layer.4.v_cache: torch.Size([1, 8, 67, 128]) - layer.4.output: torch.Size([1, 67, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02769006 9.89530284 - layer.0.v_cache 0.00000028 0.00025254 - layer.1.k_cache 0.00368886 1.00330501 - layer.1.v_cache 0.00000112 0.00091543 - layer.2.k_cache 0.00116499 0.45623751 - layer.2.v_cache 0.00000120 0.00125402 - layer.3.k_cache 0.00139456 0.53311220 - layer.3.v_cache 0.00000232 0.00212616 - layer.4.k_cache 0.00321227 0.96814124 - layer.4.v_cache 0.00000307 0.00332862 - layer.4.output 0.00021890 0.09881820 - ------------------------------------------------------------------------------------- - TOTAL 0.00271674 0.94708917 - (elements=960,512) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 960512 -Total Bytes 213168 -BPFP 1.7755 bits/point -EBPFP 3.5509 equivalent bits/point -MSE 0.947089 ----------------------- -------------------------------------------------------- -Time: 0.492s Load: 0.004s, Pack+Encode: 0.203s, Decode+Unpack: 0.284s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 67, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 67, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9471 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample663-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample663-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample68-layer4-item1.zst (83/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample68-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 123, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 123, 128) -Output shape: (1, 123, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.0.v_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.1.k_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.1.v_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.2.k_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.2.v_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.3.k_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.3.v_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.4.k_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.4.v_cache: torch.Size([1, 8, 123, 128]) -> torch.Size([1, 1, 123, 1024]) - layer.4.output: torch.Size([1, 123, 4096]) -> torch.Size([1, 1, 123, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 12,180B, BPFP=0.7736 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,704B, BPFP=1.8867 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 21,552B, BPFP=1.3689 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,120B, BPFP=1.9766 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,944B, BPFP=1.5208 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,736B, BPFP=2.0158 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,868B, BPFP=1.5795 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 31,572B, BPFP=2.0053 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 22,140B, BPFP=1.4062 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,968B, BPFP=2.0305 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 94,756B, BPFP=1.5046 -⌛️ [2/4] FRONTEND: Frontend time: 0.207s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 123, 128]) - layer.0.v_cache: torch.Size([1, 8, 123, 128]) - layer.1.k_cache: torch.Size([1, 8, 123, 128]) - layer.1.v_cache: torch.Size([1, 8, 123, 128]) - layer.2.k_cache: torch.Size([1, 8, 123, 128]) - layer.2.v_cache: torch.Size([1, 8, 123, 128]) - layer.3.k_cache: torch.Size([1, 8, 123, 128]) - layer.3.v_cache: torch.Size([1, 8, 123, 128]) - layer.4.k_cache: torch.Size([1, 8, 123, 128]) - layer.4.v_cache: torch.Size([1, 8, 123, 128]) - layer.4.output: torch.Size([1, 123, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.292s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 123, 128]) - layer.0.v_cache: torch.Size([1, 8, 123, 128]) - layer.1.k_cache: torch.Size([1, 8, 123, 128]) - layer.1.v_cache: torch.Size([1, 8, 123, 128]) - layer.2.k_cache: torch.Size([1, 8, 123, 128]) - layer.2.v_cache: torch.Size([1, 8, 123, 128]) - layer.3.k_cache: torch.Size([1, 8, 123, 128]) - layer.3.v_cache: torch.Size([1, 8, 123, 128]) - layer.4.k_cache: torch.Size([1, 8, 123, 128]) - layer.4.v_cache: torch.Size([1, 8, 123, 128]) - layer.4.output: torch.Size([1, 123, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02602563 7.93697252 - layer.0.v_cache 0.00000027 0.00022301 - layer.1.k_cache 0.00336899 0.73301641 - layer.1.v_cache 0.00000079 0.00076811 - layer.2.k_cache 0.00113470 0.39257378 - layer.2.v_cache 0.00000113 0.00111189 - layer.3.k_cache 0.00144113 0.46329151 - layer.3.v_cache 0.00000264 0.00188989 - layer.4.k_cache 0.00356668 0.88884431 - layer.4.v_cache 0.00000295 0.00301674 - layer.4.output 0.00019508 0.08606138 - ------------------------------------------------------------------------------------- - TOTAL 0.00259466 0.76899669 - (elements=1,763,328) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1763328 -Total Bytes 355540 -BPFP 1.6130 bits/point -EBPFP 3.2261 equivalent bits/point -MSE 0.768997 ----------------------- -------------------------------------------------------- -Time: 0.506s Load: 0.007s, Pack+Encode: 0.207s, Decode+Unpack: 0.292s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 123, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 123, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7690 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample68-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample68-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample7-layer4-item1.zst (84/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample7-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 157, 128) -Output shape: (1, 157, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.0.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.1.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.1.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.2.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.2.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.3.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.3.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.4.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.4.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) - layer.4.output: torch.Size([1, 157, 4096]) -> torch.Size([1, 1, 157, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 15,552B, BPFP=0.7739 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 38,680B, BPFP=1.9248 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 27,588B, BPFP=1.3728 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 41,608B, BPFP=2.0705 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 31,128B, BPFP=1.5490 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 42,516B, BPFP=2.1156 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 32,412B, BPFP=1.6129 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 42,104B, BPFP=2.0951 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 28,168B, BPFP=1.4017 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 42,804B, BPFP=2.1300 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 122,672B, BPFP=1.5261 -⌛️ [2/4] FRONTEND: Frontend time: 0.255s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 157, 128]) - layer.0.v_cache: torch.Size([1, 8, 157, 128]) - layer.1.k_cache: torch.Size([1, 8, 157, 128]) - layer.1.v_cache: torch.Size([1, 8, 157, 128]) - layer.2.k_cache: torch.Size([1, 8, 157, 128]) - layer.2.v_cache: torch.Size([1, 8, 157, 128]) - layer.3.k_cache: torch.Size([1, 8, 157, 128]) - layer.3.v_cache: torch.Size([1, 8, 157, 128]) - layer.4.k_cache: torch.Size([1, 8, 157, 128]) - layer.4.v_cache: torch.Size([1, 8, 157, 128]) - layer.4.output: torch.Size([1, 157, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.388s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 157, 128]) - layer.0.v_cache: torch.Size([1, 8, 157, 128]) - layer.1.k_cache: torch.Size([1, 8, 157, 128]) - layer.1.v_cache: torch.Size([1, 8, 157, 128]) - layer.2.k_cache: torch.Size([1, 8, 157, 128]) - layer.2.v_cache: torch.Size([1, 8, 157, 128]) - layer.3.k_cache: torch.Size([1, 8, 157, 128]) - layer.3.v_cache: torch.Size([1, 8, 157, 128]) - layer.4.k_cache: torch.Size([1, 8, 157, 128]) - layer.4.v_cache: torch.Size([1, 8, 157, 128]) - layer.4.output: torch.Size([1, 157, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02423319 7.78322801 - layer.0.v_cache 0.00000027 0.00022957 - layer.1.k_cache 0.00309445 0.76509711 - layer.1.v_cache 0.00000088 0.00081211 - layer.2.k_cache 0.00117908 0.39468306 - layer.2.v_cache 0.00000110 0.00113096 - layer.3.k_cache 0.00138553 0.46352255 - layer.3.v_cache 0.00000215 0.00194318 - layer.4.k_cache 0.00349636 0.84322663 - layer.4.v_cache 0.00000310 0.00311181 - layer.4.output 0.00020206 0.06632179 - ------------------------------------------------------------------------------------- - TOTAL 0.00244317 0.75159087 - (elements=2,250,752) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2250752 -Total Bytes 465232 -BPFP 1.6536 bits/point -EBPFP 3.3072 equivalent bits/point -MSE 0.751591 ----------------------- -------------------------------------------------------- -Time: 0.651s Load: 0.008s, Pack+Encode: 0.255s, Decode+Unpack: 0.388s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7516 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample7-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample7-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample736-layer4-item1.zst (85/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample736-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 75, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 75, 128) -Output shape: (1, 75, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.0.v_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.1.k_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.1.v_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.2.k_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.2.v_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.3.k_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.3.v_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.4.k_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.4.v_cache: torch.Size([1, 8, 75, 128]) -> torch.Size([1, 1, 75, 1024]) - layer.4.output: torch.Size([1, 75, 4096]) -> torch.Size([1, 1, 75, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,756B, BPFP=0.8079 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 19,660B, BPFP=2.0479 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,736B, BPFP=1.4308 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 21,308B, BPFP=2.2196 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,564B, BPFP=1.6213 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,768B, BPFP=2.2675 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 16,016B, BPFP=1.6683 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 21,384B, BPFP=2.2275 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,640B, BPFP=1.4208 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 22,036B, BPFP=2.2954 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 62,156B, BPFP=1.6186 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 75, 128]) - layer.0.v_cache: torch.Size([1, 8, 75, 128]) - layer.1.k_cache: torch.Size([1, 8, 75, 128]) - layer.1.v_cache: torch.Size([1, 8, 75, 128]) - layer.2.k_cache: torch.Size([1, 8, 75, 128]) - layer.2.v_cache: torch.Size([1, 8, 75, 128]) - layer.3.k_cache: torch.Size([1, 8, 75, 128]) - layer.3.v_cache: torch.Size([1, 8, 75, 128]) - layer.4.k_cache: torch.Size([1, 8, 75, 128]) - layer.4.v_cache: torch.Size([1, 8, 75, 128]) - layer.4.output: torch.Size([1, 75, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 75, 128]) - layer.0.v_cache: torch.Size([1, 8, 75, 128]) - layer.1.k_cache: torch.Size([1, 8, 75, 128]) - layer.1.v_cache: torch.Size([1, 8, 75, 128]) - layer.2.k_cache: torch.Size([1, 8, 75, 128]) - layer.2.v_cache: torch.Size([1, 8, 75, 128]) - layer.3.k_cache: torch.Size([1, 8, 75, 128]) - layer.3.v_cache: torch.Size([1, 8, 75, 128]) - layer.4.k_cache: torch.Size([1, 8, 75, 128]) - layer.4.v_cache: torch.Size([1, 8, 75, 128]) - layer.4.output: torch.Size([1, 75, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02489581 8.62647705 - layer.0.v_cache 0.00000028 0.00024274 - layer.1.k_cache 0.00341811 0.93421448 - layer.1.v_cache 0.00000080 0.00082503 - layer.2.k_cache 0.00110890 0.42200551 - layer.2.v_cache 0.00000112 0.00112954 - layer.3.k_cache 0.00134674 0.47649251 - layer.3.v_cache 0.00000209 0.00182859 - layer.4.k_cache 0.00330252 0.86772044 - layer.4.v_cache 0.00000296 0.00305649 - layer.4.output 0.00020017 0.09155171 - ------------------------------------------------------------------------------------- - TOTAL 0.00249143 0.83572852 - (elements=1,075,200) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1075200 -Total Bytes 235024 -BPFP 1.7487 bits/point -EBPFP 3.4974 equivalent bits/point -MSE 0.835729 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.005s, Pack+Encode: 0.205s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 75, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 75, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8357 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample736-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample736-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample74-layer4-item1.zst (86/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample74-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 122, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 122, 128) -Output shape: (1, 122, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.0.v_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.1.k_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.1.v_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.2.k_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.2.v_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.3.k_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.3.v_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.4.k_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.4.v_cache: torch.Size([1, 8, 122, 128]) -> torch.Size([1, 1, 122, 1024]) - layer.4.output: torch.Size([1, 122, 4096]) -> torch.Size([1, 1, 122, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 12,048B, BPFP=0.7715 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,808B, BPFP=1.9088 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 21,416B, BPFP=1.3714 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,564B, BPFP=2.0213 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,992B, BPFP=1.5364 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,952B, BPFP=2.0461 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,516B, BPFP=1.5699 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 31,528B, BPFP=2.0190 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 21,792B, BPFP=1.3955 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 32,116B, BPFP=2.0566 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 95,292B, BPFP=1.5256 -⌛️ [2/4] FRONTEND: Frontend time: 0.209s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 122, 128]) - layer.0.v_cache: torch.Size([1, 8, 122, 128]) - layer.1.k_cache: torch.Size([1, 8, 122, 128]) - layer.1.v_cache: torch.Size([1, 8, 122, 128]) - layer.2.k_cache: torch.Size([1, 8, 122, 128]) - layer.2.v_cache: torch.Size([1, 8, 122, 128]) - layer.3.k_cache: torch.Size([1, 8, 122, 128]) - layer.3.v_cache: torch.Size([1, 8, 122, 128]) - layer.4.k_cache: torch.Size([1, 8, 122, 128]) - layer.4.v_cache: torch.Size([1, 8, 122, 128]) - layer.4.output: torch.Size([1, 122, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.308s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 122, 128]) - layer.0.v_cache: torch.Size([1, 8, 122, 128]) - layer.1.k_cache: torch.Size([1, 8, 122, 128]) - layer.1.v_cache: torch.Size([1, 8, 122, 128]) - layer.2.k_cache: torch.Size([1, 8, 122, 128]) - layer.2.v_cache: torch.Size([1, 8, 122, 128]) - layer.3.k_cache: torch.Size([1, 8, 122, 128]) - layer.3.v_cache: torch.Size([1, 8, 122, 128]) - layer.4.k_cache: torch.Size([1, 8, 122, 128]) - layer.4.v_cache: torch.Size([1, 8, 122, 128]) - layer.4.output: torch.Size([1, 122, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02525679 7.65088641 - layer.0.v_cache 0.00000028 0.00023699 - layer.1.k_cache 0.00316206 0.68709602 - layer.1.v_cache 0.00000097 0.00084070 - layer.2.k_cache 0.00113900 0.38575441 - layer.2.v_cache 0.00000117 0.00114961 - layer.3.k_cache 0.00136642 0.43184008 - layer.3.v_cache 0.00000221 0.00184319 - layer.4.k_cache 0.00339647 0.79577049 - layer.4.v_cache 0.00000316 0.00300218 - layer.4.output 0.00022994 0.07353394 - ------------------------------------------------------------------------------------- - TOTAL 0.00251773 0.73232542 - (elements=1,748,992) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1748992 -Total Bytes 356024 -BPFP 1.6285 bits/point -EBPFP 3.2570 equivalent bits/point -MSE 0.732325 ----------------------- -------------------------------------------------------- -Time: 0.524s Load: 0.008s, Pack+Encode: 0.209s, Decode+Unpack: 0.308s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 122, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 122, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7323 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample74-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample74-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample75-layer4-item1.zst (87/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample75-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 142, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 142, 128) -Output shape: (1, 142, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.0.v_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.1.k_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.1.v_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.2.k_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.2.v_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.3.k_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.3.v_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.4.k_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.4.v_cache: torch.Size([1, 8, 142, 128]) -> torch.Size([1, 1, 142, 1024]) - layer.4.output: torch.Size([1, 142, 4096]) -> torch.Size([1, 1, 142, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 14,220B, BPFP=0.7824 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 36,660B, BPFP=2.0169 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 25,396B, BPFP=1.3972 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 39,008B, BPFP=2.1461 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 28,644B, BPFP=1.5759 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 39,636B, BPFP=2.1807 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 29,964B, BPFP=1.6485 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 39,516B, BPFP=2.1741 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 26,100B, BPFP=1.4360 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 40,388B, BPFP=2.2221 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 119,312B, BPFP=1.6411 -⌛️ [2/4] FRONTEND: Frontend time: 0.253s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 142, 128]) - layer.0.v_cache: torch.Size([1, 8, 142, 128]) - layer.1.k_cache: torch.Size([1, 8, 142, 128]) - layer.1.v_cache: torch.Size([1, 8, 142, 128]) - layer.2.k_cache: torch.Size([1, 8, 142, 128]) - layer.2.v_cache: torch.Size([1, 8, 142, 128]) - layer.3.k_cache: torch.Size([1, 8, 142, 128]) - layer.3.v_cache: torch.Size([1, 8, 142, 128]) - layer.4.k_cache: torch.Size([1, 8, 142, 128]) - layer.4.v_cache: torch.Size([1, 8, 142, 128]) - layer.4.output: torch.Size([1, 142, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.392s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 142, 128]) - layer.0.v_cache: torch.Size([1, 8, 142, 128]) - layer.1.k_cache: torch.Size([1, 8, 142, 128]) - layer.1.v_cache: torch.Size([1, 8, 142, 128]) - layer.2.k_cache: torch.Size([1, 8, 142, 128]) - layer.2.v_cache: torch.Size([1, 8, 142, 128]) - layer.3.k_cache: torch.Size([1, 8, 142, 128]) - layer.3.v_cache: torch.Size([1, 8, 142, 128]) - layer.4.k_cache: torch.Size([1, 8, 142, 128]) - layer.4.v_cache: torch.Size([1, 8, 142, 128]) - layer.4.output: torch.Size([1, 142, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02301751 8.24436929 - layer.0.v_cache 0.00000029 0.00023202 - layer.1.k_cache 0.00315333 0.73589153 - layer.1.v_cache 0.00000086 0.00078019 - layer.2.k_cache 0.00112591 0.37380519 - layer.2.v_cache 0.00000130 0.00106562 - layer.3.k_cache 0.00130425 0.44280756 - layer.3.v_cache 0.00000237 0.00193523 - layer.4.k_cache 0.00348660 0.83090188 - layer.4.v_cache 0.00000314 0.00298265 - layer.4.output 0.00018111 0.07534585 - ------------------------------------------------------------------------------------- - TOTAL 0.00234429 0.78115390 - (elements=2,035,712) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 2035712 -Total Bytes 438844 -BPFP 1.7246 bits/point -EBPFP 3.4492 equivalent bits/point -MSE 0.781154 ----------------------- -------------------------------------------------------- -Time: 0.653s Load: 0.008s, Pack+Encode: 0.253s, Decode+Unpack: 0.392s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 142, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 142, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7812 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample75-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample75-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample777-layer4-item1.zst (88/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample777-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 74, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 74, 128) -Output shape: (1, 74, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) -> torch.Size([1, 1, 74, 1024]) - layer.4.output: torch.Size([1, 74, 4096]) -> torch.Size([1, 1, 74, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,384B, BPFP=0.7796 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,976B, BPFP=2.0034 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,612B, BPFP=1.4371 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,988B, BPFP=2.2158 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 15,236B, BPFP=1.6085 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 21,628B, BPFP=2.2834 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,736B, BPFP=1.6613 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 21,296B, BPFP=2.2483 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,432B, BPFP=1.4181 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 21,836B, BPFP=2.3053 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 58,704B, BPFP=1.5494 -⌛️ [2/4] FRONTEND: Frontend time: 0.206s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.output: torch.Size([1, 74, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.285s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 74, 128]) - layer.0.v_cache: torch.Size([1, 8, 74, 128]) - layer.1.k_cache: torch.Size([1, 8, 74, 128]) - layer.1.v_cache: torch.Size([1, 8, 74, 128]) - layer.2.k_cache: torch.Size([1, 8, 74, 128]) - layer.2.v_cache: torch.Size([1, 8, 74, 128]) - layer.3.k_cache: torch.Size([1, 8, 74, 128]) - layer.3.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.k_cache: torch.Size([1, 8, 74, 128]) - layer.4.v_cache: torch.Size([1, 8, 74, 128]) - layer.4.output: torch.Size([1, 74, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02555943 10.00281751 - layer.0.v_cache 0.00000028 0.00024379 - layer.1.k_cache 0.00343224 0.91289716 - layer.1.v_cache 0.00000079 0.00083715 - layer.2.k_cache 0.00111603 0.43336925 - layer.2.v_cache 0.00000116 0.00116245 - layer.3.k_cache 0.00133047 0.47367694 - layer.3.v_cache 0.00000218 0.00182532 - layer.4.k_cache 0.00334566 0.93069200 - layer.4.v_cache 0.00000301 0.00299169 - layer.4.output 0.00024685 0.08788253 - ------------------------------------------------------------------------------------- - TOTAL 0.00255562 0.93657453 - (elements=1,060,864) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1060864 -Total Bytes 228828 -BPFP 1.7256 bits/point -EBPFP 3.4512 equivalent bits/point -MSE 0.936575 ----------------------- -------------------------------------------------------- -Time: 0.496s Load: 0.005s, Pack+Encode: 0.206s, Decode+Unpack: 0.285s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 74, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 74, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9366 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample777-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample777-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample778-layer4-item1.zst (89/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample778-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 61, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 61, 128) -Output shape: (1, 61, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.0.v_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.1.k_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.1.v_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.2.k_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.2.v_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.3.k_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.3.v_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.4.k_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.4.v_cache: torch.Size([1, 8, 61, 128]) -> torch.Size([1, 1, 61, 1024]) - layer.4.output: torch.Size([1, 61, 4096]) -> torch.Size([1, 1, 61, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,868B, BPFP=0.8796 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 15,216B, BPFP=1.9488 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 11,480B, BPFP=1.4703 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 16,200B, BPFP=2.0748 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 12,732B, BPFP=1.6306 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 16,196B, BPFP=2.0743 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 12,960B, BPFP=1.6598 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 15,968B, BPFP=2.0451 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 11,580B, BPFP=1.4831 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 16,280B, BPFP=2.0850 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 49,568B, BPFP=1.5871 -⌛️ [2/4] FRONTEND: Frontend time: 0.154s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 61, 128]) - layer.0.v_cache: torch.Size([1, 8, 61, 128]) - layer.1.k_cache: torch.Size([1, 8, 61, 128]) - layer.1.v_cache: torch.Size([1, 8, 61, 128]) - layer.2.k_cache: torch.Size([1, 8, 61, 128]) - layer.2.v_cache: torch.Size([1, 8, 61, 128]) - layer.3.k_cache: torch.Size([1, 8, 61, 128]) - layer.3.v_cache: torch.Size([1, 8, 61, 128]) - layer.4.k_cache: torch.Size([1, 8, 61, 128]) - layer.4.v_cache: torch.Size([1, 8, 61, 128]) - layer.4.output: torch.Size([1, 61, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.207s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 61, 128]) - layer.0.v_cache: torch.Size([1, 8, 61, 128]) - layer.1.k_cache: torch.Size([1, 8, 61, 128]) - layer.1.v_cache: torch.Size([1, 8, 61, 128]) - layer.2.k_cache: torch.Size([1, 8, 61, 128]) - layer.2.v_cache: torch.Size([1, 8, 61, 128]) - layer.3.k_cache: torch.Size([1, 8, 61, 128]) - layer.3.v_cache: torch.Size([1, 8, 61, 128]) - layer.4.k_cache: torch.Size([1, 8, 61, 128]) - layer.4.v_cache: torch.Size([1, 8, 61, 128]) - layer.4.output: torch.Size([1, 61, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02755715 8.18127441 - layer.0.v_cache 0.00000026 0.00022403 - layer.1.k_cache 0.00385674 0.89012859 - layer.1.v_cache 0.00000079 0.00086416 - layer.2.k_cache 0.00117960 0.41677982 - layer.2.v_cache 0.00000113 0.00124588 - layer.3.k_cache 0.00144752 0.48055067 - layer.3.v_cache 0.00000212 0.00186360 - layer.4.k_cache 0.00316257 0.87832110 - layer.4.v_cache 0.00000306 0.00317498 - layer.4.output 0.00025867 0.10508162 - ------------------------------------------------------------------------------------- - TOTAL 0.00273183 0.80533955 - (elements=874,496) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 874496 -Total Bytes 185048 -BPFP 1.6928 bits/point -EBPFP 3.3857 equivalent bits/point -MSE 0.805340 ----------------------- -------------------------------------------------------- -Time: 0.365s Load: 0.005s, Pack+Encode: 0.154s, Decode+Unpack: 0.207s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 61, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 61, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8053 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample778-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample778-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample807-layer4-item1.zst (90/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample807-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 59, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 59, 128) -Output shape: (1, 59, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.output: torch.Size([1, 59, 4096]) -> torch.Size([1, 1, 59, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,708B, BPFP=0.8882 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 14,920B, BPFP=1.9756 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 10,992B, BPFP=1.4555 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 16,376B, BPFP=2.1684 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 12,452B, BPFP=1.6488 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 16,608B, BPFP=2.1992 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 12,848B, BPFP=1.7013 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 16,540B, BPFP=2.1901 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 11,072B, BPFP=1.4661 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 16,652B, BPFP=2.2050 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 48,212B, BPFP=1.5960 -⌛️ [2/4] FRONTEND: Frontend time: 0.157s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.output: torch.Size([1, 59, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.200s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.output: torch.Size([1, 59, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.03100202 9.33213470 - layer.0.v_cache 0.00000029 0.00024667 - layer.1.k_cache 0.00372036 0.90893141 - layer.1.v_cache 0.00000085 0.00093540 - layer.2.k_cache 0.00115281 0.45708359 - layer.2.v_cache 0.00000145 0.00134526 - layer.3.k_cache 0.00140016 0.51547251 - layer.3.v_cache 0.00000257 0.00228990 - layer.4.k_cache 0.00323529 0.98244910 - layer.4.v_cache 0.00000321 0.00361207 - layer.4.output 0.00024217 0.11746099 - ------------------------------------------------------------------------------------- - TOTAL 0.00296341 0.90531033 - (elements=845,824) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 845824 -Total Bytes 183380 -BPFP 1.7345 bits/point -EBPFP 3.4689 equivalent bits/point -MSE 0.905310 ----------------------- -------------------------------------------------------- -Time: 0.362s Load: 0.005s, Pack+Encode: 0.157s, Decode+Unpack: 0.200s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 59, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9053 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample807-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample807-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample855-layer4-item1.zst (91/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample855-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 66, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 66, 128) -Output shape: (1, 66, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.0.v_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.1.k_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.1.v_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.2.k_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.2.v_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.3.k_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.3.v_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.4.k_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.4.v_cache: torch.Size([1, 8, 66, 128]) -> torch.Size([1, 1, 66, 1024]) - layer.4.output: torch.Size([1, 66, 4096]) -> torch.Size([1, 1, 66, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,652B, BPFP=0.7874 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 17,004B, BPFP=2.0128 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 12,028B, BPFP=1.4238 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 18,692B, BPFP=2.2126 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 13,596B, BPFP=1.6094 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 19,484B, BPFP=2.3063 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 14,100B, BPFP=1.6690 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 18,776B, BPFP=2.2225 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 11,832B, BPFP=1.4006 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 19,500B, BPFP=2.3082 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 51,228B, BPFP=1.5160 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 66, 128]) - layer.0.v_cache: torch.Size([1, 8, 66, 128]) - layer.1.k_cache: torch.Size([1, 8, 66, 128]) - layer.1.v_cache: torch.Size([1, 8, 66, 128]) - layer.2.k_cache: torch.Size([1, 8, 66, 128]) - layer.2.v_cache: torch.Size([1, 8, 66, 128]) - layer.3.k_cache: torch.Size([1, 8, 66, 128]) - layer.3.v_cache: torch.Size([1, 8, 66, 128]) - layer.4.k_cache: torch.Size([1, 8, 66, 128]) - layer.4.v_cache: torch.Size([1, 8, 66, 128]) - layer.4.output: torch.Size([1, 66, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.282s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 66, 128]) - layer.0.v_cache: torch.Size([1, 8, 66, 128]) - layer.1.k_cache: torch.Size([1, 8, 66, 128]) - layer.1.v_cache: torch.Size([1, 8, 66, 128]) - layer.2.k_cache: torch.Size([1, 8, 66, 128]) - layer.2.v_cache: torch.Size([1, 8, 66, 128]) - layer.3.k_cache: torch.Size([1, 8, 66, 128]) - layer.3.v_cache: torch.Size([1, 8, 66, 128]) - layer.4.k_cache: torch.Size([1, 8, 66, 128]) - layer.4.v_cache: torch.Size([1, 8, 66, 128]) - layer.4.output: torch.Size([1, 66, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02815026 9.94411029 - layer.0.v_cache 0.00000028 0.00023355 - layer.1.k_cache 0.00365239 0.99877456 - layer.1.v_cache 0.00000074 0.00082040 - layer.2.k_cache 0.00118996 0.45584231 - layer.2.v_cache 0.00000101 0.00109421 - layer.3.k_cache 0.00141080 0.50978429 - layer.3.v_cache 0.00000191 0.00173560 - layer.4.k_cache 0.00329622 0.99204219 - layer.4.v_cache 0.00000294 0.00297853 - layer.4.output 0.00019151 0.11245913 - ------------------------------------------------------------------------------------- - TOTAL 0.00274804 0.95408946 - (elements=946,176) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 946176 -Total Bytes 202892 -BPFP 1.7155 bits/point -EBPFP 3.4309 equivalent bits/point -MSE 0.954089 ----------------------- -------------------------------------------------------- -Time: 0.490s Load: 0.005s, Pack+Encode: 0.204s, Decode+Unpack: 0.282s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 66, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 66, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9541 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample855-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample855-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample859-layer4-item1.zst (92/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample859-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 73, 128) -Output shape: (1, 73, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) -> torch.Size([1, 1, 73, 1024]) - layer.4.output: torch.Size([1, 73, 4096]) -> torch.Size([1, 1, 73, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 7,116B, BPFP=0.7616 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 18,484B, BPFP=1.9782 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 13,464B, BPFP=1.4409 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 20,364B, BPFP=2.1794 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 14,892B, BPFP=1.5938 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 20,872B, BPFP=2.2337 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 15,556B, BPFP=1.6648 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 20,532B, BPFP=2.1973 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 13,228B, BPFP=1.4157 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 20,996B, BPFP=2.2470 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 59,184B, BPFP=1.5835 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.282s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 73, 128]) - layer.0.v_cache: torch.Size([1, 8, 73, 128]) - layer.1.k_cache: torch.Size([1, 8, 73, 128]) - layer.1.v_cache: torch.Size([1, 8, 73, 128]) - layer.2.k_cache: torch.Size([1, 8, 73, 128]) - layer.2.v_cache: torch.Size([1, 8, 73, 128]) - layer.3.k_cache: torch.Size([1, 8, 73, 128]) - layer.3.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.k_cache: torch.Size([1, 8, 73, 128]) - layer.4.v_cache: torch.Size([1, 8, 73, 128]) - layer.4.output: torch.Size([1, 73, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02483096 10.56975064 - layer.0.v_cache 0.00000028 0.00023350 - layer.1.k_cache 0.00355597 0.96982459 - layer.1.v_cache 0.00000080 0.00082321 - layer.2.k_cache 0.00115478 0.43748835 - layer.2.v_cache 0.00000105 0.00110547 - layer.3.k_cache 0.00135922 0.49357892 - layer.3.v_cache 0.00000200 0.00176507 - layer.4.k_cache 0.00333109 0.92438309 - layer.4.v_cache 0.00000279 0.00290241 - layer.4.output 0.00023199 0.07713174 - ------------------------------------------------------------------------------------- - TOTAL 0.00251192 0.97931301 - (elements=1,046,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1046528 -Total Bytes 224688 -BPFP 1.7176 bits/point -EBPFP 3.4352 equivalent bits/point -MSE 0.979313 ----------------------- -------------------------------------------------------- -Time: 0.491s Load: 0.005s, Pack+Encode: 0.203s, Decode+Unpack: 0.282s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 73, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 73, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9793 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample859-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample859-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample86-layer4-item1.zst (93/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample86-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 121, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 121, 128) -Output shape: (1, 121, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.0.v_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.1.k_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.1.v_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.2.k_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.2.v_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.3.k_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.3.v_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.4.k_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.4.v_cache: torch.Size([1, 8, 121, 128]) -> torch.Size([1, 1, 121, 1024]) - layer.4.output: torch.Size([1, 121, 4096]) -> torch.Size([1, 1, 121, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,460B, BPFP=0.7399 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 29,468B, BPFP=1.9026 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 20,980B, BPFP=1.3546 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 31,268B, BPFP=2.0189 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 23,492B, BPFP=1.5168 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,596B, BPFP=2.0400 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 24,188B, BPFP=1.5617 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 31,172B, BPFP=2.0127 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 21,272B, BPFP=1.3735 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,860B, BPFP=2.0571 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 93,464B, BPFP=1.5087 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 121, 128]) - layer.0.v_cache: torch.Size([1, 8, 121, 128]) - layer.1.k_cache: torch.Size([1, 8, 121, 128]) - layer.1.v_cache: torch.Size([1, 8, 121, 128]) - layer.2.k_cache: torch.Size([1, 8, 121, 128]) - layer.2.v_cache: torch.Size([1, 8, 121, 128]) - layer.3.k_cache: torch.Size([1, 8, 121, 128]) - layer.3.v_cache: torch.Size([1, 8, 121, 128]) - layer.4.k_cache: torch.Size([1, 8, 121, 128]) - layer.4.v_cache: torch.Size([1, 8, 121, 128]) - layer.4.output: torch.Size([1, 121, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.288s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 121, 128]) - layer.0.v_cache: torch.Size([1, 8, 121, 128]) - layer.1.k_cache: torch.Size([1, 8, 121, 128]) - layer.1.v_cache: torch.Size([1, 8, 121, 128]) - layer.2.k_cache: torch.Size([1, 8, 121, 128]) - layer.2.v_cache: torch.Size([1, 8, 121, 128]) - layer.3.k_cache: torch.Size([1, 8, 121, 128]) - layer.3.v_cache: torch.Size([1, 8, 121, 128]) - layer.4.k_cache: torch.Size([1, 8, 121, 128]) - layer.4.v_cache: torch.Size([1, 8, 121, 128]) - layer.4.output: torch.Size([1, 121, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02495229 8.10914498 - layer.0.v_cache 0.00000028 0.00022583 - layer.1.k_cache 0.00332455 0.70207782 - layer.1.v_cache 0.00000080 0.00080298 - layer.2.k_cache 0.00116246 0.38728578 - layer.2.v_cache 0.00000109 0.00106338 - layer.3.k_cache 0.00134746 0.43637987 - layer.3.v_cache 0.00000209 0.00177136 - layer.4.k_cache 0.00345892 0.84757094 - layer.4.v_cache 0.00000309 0.00306252 - layer.4.output 0.00018371 0.06993060 - ------------------------------------------------------------------------------------- - TOTAL 0.00249913 0.76922199 - (elements=1,734,656) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1734656 -Total Bytes 350220 -BPFP 1.6152 bits/point -EBPFP 3.2303 equivalent bits/point -MSE 0.769222 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.008s, Pack+Encode: 0.205s, Decode+Unpack: 0.288s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 121, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 121, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7692 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample86-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample86-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample91-layer4-item1.zst (94/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample91-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 113, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 113, 128) -Output shape: (1, 113, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) -> torch.Size([1, 1, 113, 1024]) - layer.4.output: torch.Size([1, 113, 4096]) -> torch.Size([1, 1, 113, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,876B, BPFP=0.7519 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,312B, BPFP=1.9574 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,832B, BPFP=1.3711 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,316B, BPFP=2.0960 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,212B, BPFP=1.5357 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,640B, BPFP=2.1184 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,088B, BPFP=1.5962 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,500B, BPFP=2.1087 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,108B, BPFP=1.3902 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,988B, BPFP=2.1424 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 86,224B, BPFP=1.4903 -⌛️ [2/4] FRONTEND: Frontend time: 0.204s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.output: torch.Size([1, 113, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.288s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 113, 128]) - layer.0.v_cache: torch.Size([1, 8, 113, 128]) - layer.1.k_cache: torch.Size([1, 8, 113, 128]) - layer.1.v_cache: torch.Size([1, 8, 113, 128]) - layer.2.k_cache: torch.Size([1, 8, 113, 128]) - layer.2.v_cache: torch.Size([1, 8, 113, 128]) - layer.3.k_cache: torch.Size([1, 8, 113, 128]) - layer.3.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.k_cache: torch.Size([1, 8, 113, 128]) - layer.4.v_cache: torch.Size([1, 8, 113, 128]) - layer.4.output: torch.Size([1, 113, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02530831 8.36221935 - layer.0.v_cache 0.00000029 0.00022469 - layer.1.k_cache 0.00330439 0.75972214 - layer.1.v_cache 0.00000076 0.00077049 - layer.2.k_cache 0.00114562 0.39177012 - layer.2.v_cache 0.00000109 0.00105300 - layer.3.k_cache 0.00136171 0.44661091 - layer.3.v_cache 0.00000227 0.00187274 - layer.4.k_cache 0.00342222 0.82051282 - layer.4.v_cache 0.00000313 0.00296052 - layer.4.output 0.00018801 0.07213680 - ------------------------------------------------------------------------------------- - TOTAL 0.00252156 0.79116171 - (elements=1,619,968) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1619968 -Total Bytes 333096 -BPFP 1.6450 bits/point -EBPFP 3.2899 equivalent bits/point -MSE 0.791162 ----------------------- -------------------------------------------------------- -Time: 0.499s Load: 0.007s, Pack+Encode: 0.204s, Decode+Unpack: 0.288s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 113, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 113, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7912 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample91-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample91-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample92-layer4-item1.zst (95/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample92-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 110, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.008s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 110, 128) -Output shape: (1, 110, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.0.v_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.1.k_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.1.v_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.2.k_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.2.v_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.3.k_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.3.v_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.4.k_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.4.v_cache: torch.Size([1, 8, 110, 128]) -> torch.Size([1, 1, 110, 1024]) - layer.4.output: torch.Size([1, 110, 4096]) -> torch.Size([1, 1, 110, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,592B, BPFP=0.7523 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 27,724B, BPFP=1.9690 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,616B, BPFP=1.3932 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 29,860B, BPFP=2.1207 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,060B, BPFP=1.5668 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,336B, BPFP=2.1545 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 22,784B, BPFP=1.6182 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 29,812B, BPFP=2.1173 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 19,712B, BPFP=1.4000 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 30,604B, BPFP=2.1736 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 86,188B, BPFP=1.5303 -⌛️ [2/4] FRONTEND: Frontend time: 0.203s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 110, 128]) - layer.0.v_cache: torch.Size([1, 8, 110, 128]) - layer.1.k_cache: torch.Size([1, 8, 110, 128]) - layer.1.v_cache: torch.Size([1, 8, 110, 128]) - layer.2.k_cache: torch.Size([1, 8, 110, 128]) - layer.2.v_cache: torch.Size([1, 8, 110, 128]) - layer.3.k_cache: torch.Size([1, 8, 110, 128]) - layer.3.v_cache: torch.Size([1, 8, 110, 128]) - layer.4.k_cache: torch.Size([1, 8, 110, 128]) - layer.4.v_cache: torch.Size([1, 8, 110, 128]) - layer.4.output: torch.Size([1, 110, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.288s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 110, 128]) - layer.0.v_cache: torch.Size([1, 8, 110, 128]) - layer.1.k_cache: torch.Size([1, 8, 110, 128]) - layer.1.v_cache: torch.Size([1, 8, 110, 128]) - layer.2.k_cache: torch.Size([1, 8, 110, 128]) - layer.2.v_cache: torch.Size([1, 8, 110, 128]) - layer.3.k_cache: torch.Size([1, 8, 110, 128]) - layer.3.v_cache: torch.Size([1, 8, 110, 128]) - layer.4.k_cache: torch.Size([1, 8, 110, 128]) - layer.4.v_cache: torch.Size([1, 8, 110, 128]) - layer.4.output: torch.Size([1, 110, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02547969 8.37513039 - layer.0.v_cache 0.00000028 0.00022689 - layer.1.k_cache 0.00323236 0.80553429 - layer.1.v_cache 0.00000082 0.00077351 - layer.2.k_cache 0.00114874 0.40668002 - layer.2.v_cache 0.00000131 0.00111505 - layer.3.k_cache 0.00138397 0.45871370 - layer.3.v_cache 0.00000212 0.00180291 - layer.4.k_cache 0.00346958 0.86177174 - layer.4.v_cache 0.00000311 0.00305064 - layer.4.output 0.00021756 0.08249486 - ------------------------------------------------------------------------------------- - TOTAL 0.00254230 0.80319847 - (elements=1,576,960) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1576960 -Total Bytes 329288 -BPFP 1.6705 bits/point -EBPFP 3.3410 equivalent bits/point -MSE 0.803198 ----------------------- -------------------------------------------------------- -Time: 0.499s Load: 0.008s, Pack+Encode: 0.203s, Decode+Unpack: 0.288s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 110, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 110, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8032 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample92-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample92-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample925-layer4-item1.zst (96/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample925-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 65, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 65, 128) -Output shape: (1, 65, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.0.v_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.1.k_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.1.v_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.2.k_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.2.v_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.3.k_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.3.v_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.4.k_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.4.v_cache: torch.Size([1, 8, 65, 128]) -> torch.Size([1, 1, 65, 1024]) - layer.4.output: torch.Size([1, 65, 4096]) -> torch.Size([1, 1, 65, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,640B, BPFP=0.7981 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 16,300B, BPFP=1.9591 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 11,580B, BPFP=1.3918 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 18,076B, BPFP=2.1726 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 13,040B, BPFP=1.5673 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 18,636B, BPFP=2.2399 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 13,944B, BPFP=1.6760 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 18,812B, BPFP=2.2611 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 11,652B, BPFP=1.4005 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 19,104B, BPFP=2.2962 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 50,356B, BPFP=1.5131 -⌛️ [2/4] FRONTEND: Frontend time: 0.201s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 65, 128]) - layer.0.v_cache: torch.Size([1, 8, 65, 128]) - layer.1.k_cache: torch.Size([1, 8, 65, 128]) - layer.1.v_cache: torch.Size([1, 8, 65, 128]) - layer.2.k_cache: torch.Size([1, 8, 65, 128]) - layer.2.v_cache: torch.Size([1, 8, 65, 128]) - layer.3.k_cache: torch.Size([1, 8, 65, 128]) - layer.3.v_cache: torch.Size([1, 8, 65, 128]) - layer.4.k_cache: torch.Size([1, 8, 65, 128]) - layer.4.v_cache: torch.Size([1, 8, 65, 128]) - layer.4.output: torch.Size([1, 65, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.282s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 65, 128]) - layer.0.v_cache: torch.Size([1, 8, 65, 128]) - layer.1.k_cache: torch.Size([1, 8, 65, 128]) - layer.1.v_cache: torch.Size([1, 8, 65, 128]) - layer.2.k_cache: torch.Size([1, 8, 65, 128]) - layer.2.v_cache: torch.Size([1, 8, 65, 128]) - layer.3.k_cache: torch.Size([1, 8, 65, 128]) - layer.3.v_cache: torch.Size([1, 8, 65, 128]) - layer.4.k_cache: torch.Size([1, 8, 65, 128]) - layer.4.v_cache: torch.Size([1, 8, 65, 128]) - layer.4.output: torch.Size([1, 65, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02613364 10.12844614 - layer.0.v_cache 0.00000028 0.00024876 - layer.1.k_cache 0.00351263 0.90804267 - layer.1.v_cache 0.00000086 0.00083444 - layer.2.k_cache 0.00113140 0.43320465 - layer.2.v_cache 0.00000103 0.00110171 - layer.3.k_cache 0.00134002 0.48046540 - layer.3.v_cache 0.00000212 0.00186008 - layer.4.k_cache 0.00331905 0.88624408 - layer.4.v_cache 0.00000293 0.00305120 - layer.4.output 0.00019456 0.09674199 - ------------------------------------------------------------------------------------- - TOTAL 0.00258730 0.94503336 - (elements=931,840) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 931840 -Total Bytes 198140 -BPFP 1.7011 bits/point -EBPFP 3.4021 equivalent bits/point -MSE 0.945033 ----------------------- -------------------------------------------------------- -Time: 0.489s Load: 0.005s, Pack+Encode: 0.201s, Decode+Unpack: 0.282s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 65, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 65, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9450 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample925-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample925-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample95-layer4-item1.zst (97/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample95-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 116, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.007s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 116, 128) -Output shape: (1, 116, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.0.v_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.1.k_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.1.v_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.2.k_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.2.v_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.3.k_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.3.v_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.4.k_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.4.v_cache: torch.Size([1, 8, 116, 128]) -> torch.Size([1, 1, 116, 1024]) - layer.4.output: torch.Size([1, 116, 4096]) -> torch.Size([1, 1, 116, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 11,232B, BPFP=0.7565 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,772B, BPFP=1.9378 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 20,156B, BPFP=1.3575 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,816B, BPFP=2.0754 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,792B, BPFP=1.5350 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 31,012B, BPFP=2.0886 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,400B, BPFP=1.5760 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,640B, BPFP=2.0636 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,448B, BPFP=1.3772 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,392B, BPFP=2.1142 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 90,796B, BPFP=1.5288 -⌛️ [2/4] FRONTEND: Frontend time: 0.205s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 116, 128]) - layer.0.v_cache: torch.Size([1, 8, 116, 128]) - layer.1.k_cache: torch.Size([1, 8, 116, 128]) - layer.1.v_cache: torch.Size([1, 8, 116, 128]) - layer.2.k_cache: torch.Size([1, 8, 116, 128]) - layer.2.v_cache: torch.Size([1, 8, 116, 128]) - layer.3.k_cache: torch.Size([1, 8, 116, 128]) - layer.3.v_cache: torch.Size([1, 8, 116, 128]) - layer.4.k_cache: torch.Size([1, 8, 116, 128]) - layer.4.v_cache: torch.Size([1, 8, 116, 128]) - layer.4.output: torch.Size([1, 116, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.289s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 116, 128]) - layer.0.v_cache: torch.Size([1, 8, 116, 128]) - layer.1.k_cache: torch.Size([1, 8, 116, 128]) - layer.1.v_cache: torch.Size([1, 8, 116, 128]) - layer.2.k_cache: torch.Size([1, 8, 116, 128]) - layer.2.v_cache: torch.Size([1, 8, 116, 128]) - layer.3.k_cache: torch.Size([1, 8, 116, 128]) - layer.3.v_cache: torch.Size([1, 8, 116, 128]) - layer.4.k_cache: torch.Size([1, 8, 116, 128]) - layer.4.v_cache: torch.Size([1, 8, 116, 128]) - layer.4.output: torch.Size([1, 116, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02494503 8.38237736 - layer.0.v_cache 0.00000027 0.00023507 - layer.1.k_cache 0.00326550 0.76380144 - layer.1.v_cache 0.00000084 0.00082347 - layer.2.k_cache 0.00114506 0.40341486 - layer.2.v_cache 0.00000107 0.00109943 - layer.3.k_cache 0.00140586 0.46438592 - layer.3.v_cache 0.00000237 0.00191058 - layer.4.k_cache 0.00329219 0.81805736 - layer.4.v_cache 0.00000320 0.00315066 - layer.4.output 0.00021339 0.07859512 - ------------------------------------------------------------------------------------- - TOTAL 0.00249392 0.79668833 - (elements=1,662,976) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1662976 -Total Bytes 341456 -BPFP 1.6426 bits/point -EBPFP 3.2853 equivalent bits/point -MSE 0.796688 ----------------------- -------------------------------------------------------- -Time: 0.501s Load: 0.007s, Pack+Encode: 0.205s, Decode+Unpack: 0.289s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 116, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 116, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7967 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample95-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample95-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample967-layer4-item1.zst (98/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample967-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 59, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.005s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 59, 128) -Output shape: (1, 59, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) -> torch.Size([1, 1, 59, 1024]) - layer.4.output: torch.Size([1, 59, 4096]) -> torch.Size([1, 1, 59, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,276B, BPFP=0.8310 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 14,580B, BPFP=1.9306 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 10,792B, BPFP=1.4290 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 16,332B, BPFP=2.1626 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 12,056B, BPFP=1.5964 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 16,448B, BPFP=2.1780 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 12,476B, BPFP=1.6520 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 16,188B, BPFP=2.1435 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 10,836B, BPFP=1.4349 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 16,332B, BPFP=2.1626 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 47,108B, BPFP=1.5595 -⌛️ [2/4] FRONTEND: Frontend time: 0.153s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.output: torch.Size([1, 59, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.200s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 59, 128]) - layer.0.v_cache: torch.Size([1, 8, 59, 128]) - layer.1.k_cache: torch.Size([1, 8, 59, 128]) - layer.1.v_cache: torch.Size([1, 8, 59, 128]) - layer.2.k_cache: torch.Size([1, 8, 59, 128]) - layer.2.v_cache: torch.Size([1, 8, 59, 128]) - layer.3.k_cache: torch.Size([1, 8, 59, 128]) - layer.3.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.k_cache: torch.Size([1, 8, 59, 128]) - layer.4.v_cache: torch.Size([1, 8, 59, 128]) - layer.4.output: torch.Size([1, 59, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02763205 10.07965502 - layer.0.v_cache 0.00000028 0.00025296 - layer.1.k_cache 0.00373147 0.92043298 - layer.1.v_cache 0.00000088 0.00085972 - layer.2.k_cache 0.00131257 0.42840376 - layer.2.v_cache 0.00000108 0.00114188 - layer.3.k_cache 0.00142120 0.48627950 - layer.3.v_cache 0.00000202 0.00176378 - layer.4.k_cache 0.00321287 0.91481160 - layer.4.v_cache 0.00000283 0.00289157 - layer.4.output 0.00027608 0.11429859 - ------------------------------------------------------------------------------------- - TOTAL 0.00274440 0.94954908 - (elements=845,824) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 845824 -Total Bytes 179424 -BPFP 1.6970 bits/point -EBPFP 3.3941 equivalent bits/point -MSE 0.949549 ----------------------- -------------------------------------------------------- -Time: 0.358s Load: 0.005s, Pack+Encode: 0.153s, Decode+Unpack: 0.200s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 59, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 59, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.9495 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample967-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample967-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample969-layer4-item1.zst (99/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample969-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 60, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.004s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 60, 128) -Output shape: (1, 60, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.0.v_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.1.k_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.1.v_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.2.k_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.2.v_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.3.k_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.3.v_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.4.k_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.4.v_cache: torch.Size([1, 8, 60, 128]) -> torch.Size([1, 1, 60, 1024]) - layer.4.output: torch.Size([1, 60, 4096]) -> torch.Size([1, 1, 60, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 6,368B, BPFP=0.8292 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 15,004B, BPFP=1.9536 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 11,124B, BPFP=1.4484 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 16,292B, BPFP=2.1214 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 12,680B, BPFP=1.6510 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 16,496B, BPFP=2.1479 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 12,812B, BPFP=1.6682 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 16,168B, BPFP=2.1052 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 11,184B, BPFP=1.4563 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 16,396B, BPFP=2.1349 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 48,600B, BPFP=1.5820 -⌛️ [2/4] FRONTEND: Frontend time: 0.152s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 60, 128]) - layer.0.v_cache: torch.Size([1, 8, 60, 128]) - layer.1.k_cache: torch.Size([1, 8, 60, 128]) - layer.1.v_cache: torch.Size([1, 8, 60, 128]) - layer.2.k_cache: torch.Size([1, 8, 60, 128]) - layer.2.v_cache: torch.Size([1, 8, 60, 128]) - layer.3.k_cache: torch.Size([1, 8, 60, 128]) - layer.3.v_cache: torch.Size([1, 8, 60, 128]) - layer.4.k_cache: torch.Size([1, 8, 60, 128]) - layer.4.v_cache: torch.Size([1, 8, 60, 128]) - layer.4.output: torch.Size([1, 60, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.198s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 60, 128]) - layer.0.v_cache: torch.Size([1, 8, 60, 128]) - layer.1.k_cache: torch.Size([1, 8, 60, 128]) - layer.1.v_cache: torch.Size([1, 8, 60, 128]) - layer.2.k_cache: torch.Size([1, 8, 60, 128]) - layer.2.v_cache: torch.Size([1, 8, 60, 128]) - layer.3.k_cache: torch.Size([1, 8, 60, 128]) - layer.3.v_cache: torch.Size([1, 8, 60, 128]) - layer.4.k_cache: torch.Size([1, 8, 60, 128]) - layer.4.v_cache: torch.Size([1, 8, 60, 128]) - layer.4.output: torch.Size([1, 60, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02844323 8.41371663 - layer.0.v_cache 0.00000027 0.00023504 - layer.1.k_cache 0.00360089 0.83916245 - layer.1.v_cache 0.00000087 0.00089474 - layer.2.k_cache 0.00116909 0.43694553 - layer.2.v_cache 0.00000119 0.00125656 - layer.3.k_cache 0.00139808 0.48481585 - layer.3.v_cache 0.00000212 0.00191595 - layer.4.k_cache 0.00324009 0.90498645 - layer.4.v_cache 0.00000310 0.00319181 - layer.4.output 0.00020032 0.10127386 - ------------------------------------------------------------------------------------- - TOTAL 0.00276144 0.82087260 - (elements=860,160) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 860160 -Total Bytes 183124 -BPFP 1.7032 bits/point -EBPFP 3.4063 equivalent bits/point -MSE 0.820873 ----------------------- -------------------------------------------------------- -Time: 0.354s Load: 0.004s, Pack+Encode: 0.152s, Decode+Unpack: 0.198s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 60, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 60, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8209 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample969-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample969-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample97-layer4-item1.zst (100/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample97-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 114, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.006s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 114, 128) -Output shape: (1, 114, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.0.v_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.1.k_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.1.v_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.2.k_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.2.v_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.3.k_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.3.v_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.4.k_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.4.v_cache: torch.Size([1, 8, 114, 128]) -> torch.Size([1, 1, 114, 1024]) - layer.4.output: torch.Size([1, 114, 4096]) -> torch.Size([1, 1, 114, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 10,688B, BPFP=0.7325 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 28,320B, BPFP=1.9408 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 19,916B, BPFP=1.3649 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 30,396B, BPFP=2.0831 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 22,380B, BPFP=1.5337 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 30,864B, BPFP=2.1151 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 23,232B, BPFP=1.5921 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 30,300B, BPFP=2.0765 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 20,204B, BPFP=1.3846 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 31,052B, BPFP=2.1280 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 88,144B, BPFP=1.5101 -⌛️ [2/4] FRONTEND: Frontend time: 0.216s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 114, 128]) - layer.0.v_cache: torch.Size([1, 8, 114, 128]) - layer.1.k_cache: torch.Size([1, 8, 114, 128]) - layer.1.v_cache: torch.Size([1, 8, 114, 128]) - layer.2.k_cache: torch.Size([1, 8, 114, 128]) - layer.2.v_cache: torch.Size([1, 8, 114, 128]) - layer.3.k_cache: torch.Size([1, 8, 114, 128]) - layer.3.v_cache: torch.Size([1, 8, 114, 128]) - layer.4.k_cache: torch.Size([1, 8, 114, 128]) - layer.4.v_cache: torch.Size([1, 8, 114, 128]) - layer.4.output: torch.Size([1, 114, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.296s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 114, 128]) - layer.0.v_cache: torch.Size([1, 8, 114, 128]) - layer.1.k_cache: torch.Size([1, 8, 114, 128]) - layer.1.v_cache: torch.Size([1, 8, 114, 128]) - layer.2.k_cache: torch.Size([1, 8, 114, 128]) - layer.2.v_cache: torch.Size([1, 8, 114, 128]) - layer.3.k_cache: torch.Size([1, 8, 114, 128]) - layer.3.v_cache: torch.Size([1, 8, 114, 128]) - layer.4.k_cache: torch.Size([1, 8, 114, 128]) - layer.4.v_cache: torch.Size([1, 8, 114, 128]) - layer.4.output: torch.Size([1, 114, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02435601 8.40135541 - layer.0.v_cache 0.00000028 0.00022643 - layer.1.k_cache 0.00337316 0.75997911 - layer.1.v_cache 0.00000076 0.00080010 - layer.2.k_cache 0.00114524 0.40666272 - layer.2.v_cache 0.00000110 0.00115780 - layer.3.k_cache 0.00139029 0.47045935 - layer.3.v_cache 0.00000222 0.00187027 - layer.4.k_cache 0.00339712 0.85309514 - layer.4.v_cache 0.00000302 0.00316253 - layer.4.output 0.00021100 0.08482022 - ------------------------------------------------------------------------------------- - TOTAL 0.00246523 0.80271784 - (elements=1,634,304) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 1634304 -Total Bytes 335496 -BPFP 1.6423 bits/point -EBPFP 3.2845 equivalent bits/point -MSE 0.802718 ----------------------- -------------------------------------------------------- -Time: 0.517s Load: 0.006s, Pack+Encode: 0.216s, Decode+Unpack: 0.296s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 114, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 114, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.8027 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_gsm8k/sample97-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample97-layer4-item1.zst ------------------------- ---------------------------- -TOTAL PROCESSING SUMMARY ------------------------- ---------------------------- -Total files 100 -Avg BPFP 1.6907 bits/point -Avg EBPFP 3.3814 equivalent bits/point -Avg MSE 0.863113 -Avg Time 0.516s ------------------------- ---------------------------- +version https://git-lfs.github.com/spec/v1 +oid sha256:b63ec939aa8efd118697d7bf4be066c24b1b4c338895e733318f2b2955bbf770 +size 1119009 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..9d080027a191fcb488d7ecd9bcf4cf14f93ba5c6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef6383d1c36fee04215b16073b37057d90f4efd934d0826989880a18b63f2439 +size 3034413 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample100-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample100-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..24874cbf5352e08753f92a240e111d85b9a014f9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample100-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20f04c6233049273f5c98731e74aee48007deff83c4600361a1b67819210ac68 +size 1931176 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample108-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample108-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f5f9b367b9e3cd649bbc1bf765e6330f07843330 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample108-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:372773c1bf7b0b50c610835f4ee918f5a304a352b0b87aae456031c59ce17bbc +size 1726050 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1087-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1087-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f307e4d1645d24387611dd3e02a1a561c95656df --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1087-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f23f78ec0508010ef46ecb2497767c9404f2a48f9d84a46a750cfef8c00e4d2 +size 899498 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1128-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1128-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4290e661ac0bfbfd016690d5c1db98d77dc21ff0 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1128-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7324a8859b1994dc7fc7e90eb1cbf533f8f3dd571543675b4475655e9b60fbe1 +size 1024175 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample117-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample117-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2a4547fcb33b2edaabb5e9cdd01d118a61c2d546 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample117-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:931c9ecb3939c45c8de918d668222c15375859ef1fe64d903649ee2a7107ea22 +size 2010379 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample120-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample120-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..088cbd30b05d2c506aa10fc8a3c23643e0b20ed1 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample120-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22baae19229387cafed61c2eaf181f39f3421497e537d3328588be31337d96fc +size 1907850 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1295-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1295-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b6db3b29805404779d487165d7c1129926c98cc9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample1295-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53d1f2f546654afd984e2fc16ef2744385498a7c05e518c7e5151528e5cfdee5 +size 848235 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample13-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample13-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5419685a5679fb44f9f2468f3cd4691c85e0d5f6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample13-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be5b8f9fcaaf9debc7be08bb442c3e4089cec3552724a651e54cb73585d57a25 +size 2726291 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample130-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample130-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..022e08a4116d661b4f2100fd9cbb7bd577b6c691 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample130-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5e64e44b2eeb3d89a79dd13a2a72d791483d1ddd578dae88d3a921d00f866f2 +size 1781295 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample144-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample144-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d5a7a58fbae7081746028d01609460d121fa1ea0 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample144-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0710de3fee9fff91ef09eef4208c7b7a091b71e903962aea1417775a3bdc23e +size 1689076 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample145-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample145-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d2a206ba6d9c57781e9132438f27af77d3de9105 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample145-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bb4f148632dc9ca159f8d0f46bd1317935cf3c6faf9e61ae88b4f133a0e3ba7 +size 1714735 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample146-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample146-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c2193a9a6a33a1044245cac891a7e7deaa75dd74 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample146-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3016497cfb48343fe6e54ddc91fa90e888c2629419b6a1fc9ba0b33b5bc5de4 +size 1751382 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample147-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample147-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1232b282abaffaf18d2e400ffc73a067ee3b728c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample147-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2babba783d50995d14479f527f065bf5f7ea697f7ddb0a48feb12ebaa3dbc64a +size 1743028 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample150-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample150-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f7e71a36aaa1f141d7f5fdd4e946ef43888c266a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample150-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43b3c177565ea682909556d5e18ae59d5579311d2d176b2601ace013dcbc1d87 +size 1774748 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample153-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample153-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..34caa0fb0ab4628632cf15e9ec188905772da6f8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample153-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0db4ac30f22d02c669adfb032db42c1f77d938bb6861cfe2415544365667d8b +size 1721430 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample154-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample154-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..dbcb68d0149bf6dbf20f85ec6efb3733195d043c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample154-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e161a922953ccdf7a490449627f67f1c3c97ce1cd9cab53bcff01699d935a2d +size 1578463 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample157-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample157-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..bb6b1c617081f7f95f6d9f4bfcd65b910d691942 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample157-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8355507145ed824eb88dd0d78c54d99f63b0e96b97a72b23f32577388a31c67 +size 1857298 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample159-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample159-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..70bfd1921c11cd10fc5af60e47daa0a9f33a6253 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample159-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd836b3ff7713e5e8831ca0bb53a75f778dc3613eb2d33e6755b8399b88bc275 +size 1449521 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample165-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample165-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..03e4cff8d6ebca61f1a9b0338e62f9de9c0e291c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample165-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91464181df2d60fc2db3645e301b9896d961416c35113a3d372b0186de6b4a17 +size 1930395 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample167-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample167-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0c2175d66979e2117ba5bf717ae5521a96711911 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample167-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6242be15124eb070b3f67417ff0628b10326d124605b9ec9e699b80217a3de4 +size 1696721 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample168-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample168-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d0e45884a3df891df1f4f25096dd62ccc3f50cdd --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample168-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:412573d55653a2e7ca3c5a19ae742d46b099ac556715e05b8ad057296e41a158 +size 1828150 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample174-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample174-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2d38a14e5cc5c636e1a5a81ff7e2b9320659faf8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample174-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af5256cc92b880b80cdc6ebb6f6c0fa2e2bdb2afff274e3ac65cab49edd030ee +size 1831413 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample182-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample182-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..acb613d9fd131dfeb0c9f9aaebd345cedc62001d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample182-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:986b292ee8bba75d883bf3b5a5943a568ec98991b5d4d300378175c8795bc8f5 +size 1642733 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample185-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample185-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2654241ac6d405528fca8b8871c79de915b5efb8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample185-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf924a2cdb0aa1cfcfe71a7595a0ddc5de53c99d78de63427460c25e5f206232 +size 1707425 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample191-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample191-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ce7f93385f9d495817fb665e342907703eddcb1d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample191-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:035f91830aa194acfd0b4572aa3415ff52d796deb8ff08d2ef20695a35449265 +size 1656946 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample196-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample196-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..9187876a138b08b212ba3be68dd53a11f6fc0fbb --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample196-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb276a9de7d2975c7c20d8f445ab2ce7ec832a3d93f42b715d75a1e9c8f70c2f +size 1713552 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample197-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample197-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b53acd8fc5cc87bbefc61917ea6b5f5657450ad2 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample197-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f92c969c860be59986eebcbad316d83eb6d0c30e33fc9b83bad25bd8cdc1245c +size 1639896 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample2-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample2-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4446ced31c301af2eee1d4542dd2bc090b61e3e6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample2-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e744487274bc2ff4d52a254939e2769bbbc89c806f2cba61f24fdc534f8e5d1b +size 3126689 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample201-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample201-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..bd3eafae0d4da80776636349857c55f4dd1aa1dc --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample201-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11eef509dfa68a05dd3ceef6efe9ce0b97aaf260f6fa3ad2d0b7d22866064150 +size 1684950 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample212-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample212-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..db6caa5b0a9e279f39f4801afcb5a9164bf8c253 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample212-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca770554215f42a1255533ac9666c7f40c76e0ff0ff50ddd20d0eac6f1a27397 +size 1669541 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample214-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample214-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..12e2158a6063d95152be48f2ec3cefb65a209110 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample214-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28a45b0a043df90f464dde1ff68311bdb3d742b656b0353e8018deeb7510538b +size 1757630 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample224-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample224-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..a37bc9be6d1d71942191c6408ab2b13d3ed25f21 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample224-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad8a86e4f189d3f0c772315daacbbb5b1e04de6937f507172dde720af3299559 +size 1823656 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample227-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample227-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3be337a708360b8684cb4b09e551467888fe0237 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample227-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ebf9282a51e1600cdd4d0718dd2e2dd39a500959f976a4a9773ec4e71aa75cd +size 1714184 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample233-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample233-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f3d41ed2ddc04bf168a69577ee992d2ba9ab7762 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample233-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6449302afd03666cd440c73624c905f59a1d3e1721207afe039f20c24103f84b +size 1520941 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample241-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample241-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..df1b95dcf08c8592d3edb50e4bcdde0d7cedbb86 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample241-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57fcb56cc64d842e037e23cbc4674f5f09cb20c2df73cac407aebd701f88fcda +size 1720239 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample250-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample250-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..dc3bb23403ce2a85a984ee52003cddde7f1f06be --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample250-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39945e5ef419c29830ea796ccadfe16d6cfd73b80c878b1f0f27f33efca6825c +size 1738362 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample251-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample251-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..37ed2ff1a09ffb2951e82a5eaa2b0793557a0204 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample251-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff901c4fb31425789d6007285716abb554f5d6ad52f7f326c7625b9801ee0950 +size 1753643 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample257-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample257-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0180efb2cf06ea75264df40bd5abc2984f07fd9f --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample257-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bc95b276ee8cc0bf2ff3c5e1f47faf520224f975bf068b037fbbc65339a1b43 +size 1538872 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample258-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample258-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..db29944a2df928d1d33854a29b00f5d45598f7ea --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample258-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f34c8d1757fa1ec643819e1e99448c1ac3b0271bf7d98721d17d79c4eb732726 +size 1588391 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample263-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample263-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..429b7a0218b781b3d80e1af09b755d8c5355b2e8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample263-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5522a13c33bf443fe87fe48cfba6b404e03cfad9bf32c672fc42278acd7a32fc +size 1591409 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample274-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample274-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c8fe02ee13204bef76699b7e99bf123485c54efe --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample274-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5f66b8d6ed16497152f4c1f8d5f2a008ccf3e111dad6678397c93f2abf12626 +size 1628622 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample282-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample282-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..6ef1cb45457ea775d7b17c451e1b47c4a2abf74b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample282-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ead6ee2462bf0ebe6238aff111b7273812a2191960001b6365aa242ed74863c1 +size 1548080 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample290-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample290-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2df8e18ac872d8aea1905c5678856a9f43c2e81c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample290-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ac132a2701205361557c2a23658e4bae4af149981e171c5a0b1d86414c0ac10 +size 1667814 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample3-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample3-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3e8f1789ed021046de1f63fc5b1d7c38c468f500 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample3-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20dbd4da8f8d08bc1423c299c3731d6e235ffa9896f6d2056ec6519187a37937 +size 2640517 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample307-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample307-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d47ca961e5e46baf579624071027ddbd63363fd0 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample307-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f5c54e0fbb61a9c0169ad9845fdee30b53b6678e86c93b888ad3a49e5192b2c +size 1657859 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample313-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample313-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e145f9320de0c7635fbc3b3f2baf10cbc8f70260 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample313-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3abec0a12725942f934d62b7c829881d2fa2677c5ea9480f8632cbd9607317f9 +size 1444072 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample319-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample319-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2d8d9fbb8f842a9b1680e8453214ee226a64aa66 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample319-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6f8fb55b229057ae47a19a7c6bd2f52f576092330d6c16ee584f72a4e42551c +size 1648363 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample333-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample333-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b914fc8dbb85bf58c2818a03a9cabdab6b8818c1 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample333-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25c6fbc72e5fb86cea6748373c5f420a06cadb34531dd141347a42bf1ea6f43d +size 1596641 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample36-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample36-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ed581bc78203759fb4f03a9cce4fe0cb99346a94 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample36-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6616ea0b09fd659153061e14f28264c637f72611e72bfde4b2631a02502ab1c8 +size 2117539 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample365-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample365-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d57ab4a2289fcba7770acb6908e3a6c82f00bc79 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample365-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3070cc2f2e9093b8f4a8752a01683c7476c35b583daa3421b0c91cb8845ef1bd +size 1454949 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample38-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample38-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..05e9b4d818fe08ab03a0c1a03c5129cbf753afc3 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample38-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a96829cd7effeee02104458e67729847f22c6765920da13bed6295c421c186d9 +size 2197938 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample388-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample388-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e8440ebb8fbf44f899a0c06d765696af75cf7a2a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample388-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:28f8f860fcb04bc451587a8c16c0b7cad783bc17aacdfb359ff684fd2b39706d +size 1502751 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample390-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample390-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5a9c21c3dfe11c70fe174a68ae139bcce1a54fa4 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample390-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d8f712be26388ddbfd9d82a8a19ee7ea2190acf8916753ae758a18191fc9451 +size 1321954 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample40-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample40-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..eb63d36ca2410ce91667f0aaf9c18ea227685b3a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample40-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20c24455a351ec66b7a7d460f35edcf772cd4e2bdf3d2d4eff731f6d639b7a33 +size 2223121 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample412-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample412-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7cfaa0a9d388c652aa62d81c8481bc020f436123 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample412-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9944518291ae76aab82803a39c935652a8f4b23cca83c0c779e0c1724b17e2a +size 1349421 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample414-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample414-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..82c09c03cc8012c769ee15661131710871fbaa44 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample414-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b151e7f26169566b5a92eed136cab6cdc9835fd00c75f35c41b01d207014f37 +size 1512637 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample42-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample42-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ba7018e3738db03e29f49787ea29b9cd31dd4152 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample42-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5ff6accf54041accf64b8b469f3cb980867d4b0bcf2c5420601d391b8736509 +size 2347926 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample44-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample44-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b73c8e8982e3d7c7645261613e032038da96c2bf --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample44-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc333c97430ee9e7b6a8872dbc33401ab123128c31cada03cc7657bc967eb24d +size 2336299 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample454-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample454-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ddeb17abbd9427199e5bd27b78b7030e1efbea3b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample454-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b28f3c45aa54a444f2f3f8189a9d75ce3032a3df4a394dc89c5694c2656d7875 +size 1379559 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample464-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample464-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..147dc0f6212f4ea1d7573b99795c7caea65ceb59 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample464-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e733ef5acbcebcfa1abae548435f7ca3de09cf25b2ac21c5e523abd68c17a517 +size 1670306 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample478-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample478-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0760e89698b39580785b34b12c652efabc0a2b94 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample478-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39c0d4e0c8356a7d954c2c43add2fb37662092a6f2bff3344d98440af34b2954 +size 1367233 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample48-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample48-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4301f857e072b32e904fce5977808c976f45a6c4 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample48-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b04895144e2664ca4470cefedc918142aa82bc2b9955f54300765458749e044 +size 1971951 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample485-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample485-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..8685e40f320e6829aaa133b5d5d4306fc52f304d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample485-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c180b21ab03e5e94b03c5d43e03722e5103e9bb8a3a6f5f79ce25a0e74b7fec1 +size 1395684 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample487-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample487-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3c37a38df9c53428f9f40c550de6d395065c9560 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample487-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f1d3aaf581e484d1270d0508b83f603bd62467604597fe602e33d6ca1a51492 +size 1404084 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample495-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample495-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ab2cc7b68114f6acbc9af66b5ac437655f5e6b58 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample495-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4a3c94da59d9c5df650a3cfba4b408c03b12da1a58d91d053925fd4864495084 +size 1454187 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample50-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample50-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..a4f7ce873dfe08937db73a926e43e81bb40c51ff --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample50-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a012586fed1b3fd9d470fb25c1d9ebcaf89eac48763b7af3d2698b0f36c06a4d +size 2072261 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample516-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample516-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..beca90f25c8b5846c3b9d3b429b0dff4b41aa71a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample516-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b75132456257c030afde987b07e04b92a15444f8874e882c9f0e1fecbcce4767 +size 1501832 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample52-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample52-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..497a53696ce3c24f85d5729da04e3634a5cf0c87 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample52-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:adfd19d86e1799fff0132c477fec331de7d5009f7a6e83ce35f11ab0fc920bc3 +size 2347135 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample53-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample53-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5c6a6b03837fd6532a822c50cd9ef94b8e5fee1b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample53-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89dd85a6f55c9a9aa73f992208567f0ab54dad74f42f687de1ff308d54bbf2f6 +size 2204149 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample543-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample543-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5e8063c94e299bb0a7f2edf6bdb299e04e43e8c5 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample543-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b03bcd0ca7bdb2a9e699b00a8436d28f3dc6202aea88503c60865f13764554f3 +size 1270607 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample561-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample561-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d8a25337a31b8fc065c4f485e306658eef430d75 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample561-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:749e7a105b0b877e2f9a54e4986e6112b068ab2b85022c07828f2e1831218c88 +size 1223983 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample570-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample570-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1b10f467b0234c60538925970d9b9a1771b07736 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample570-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43d26c52dde2f62fd6ce349accf0e13ddc3b0d80e5efb0f19997d499d951a851 +size 1287554 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample581-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample581-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f7b10ac7df05ee1250b15864de666548f649183b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample581-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0350e815d901ef73b563cd23765972006cd267623a846c9d74492e1e53efcc57 +size 1298252 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample584-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample584-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e1aa46c4701aad1c24078758841d1aa11c47ea86 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample584-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:27dd6553f0d824ddce657ebf26e997dced74af39a6054f8fa71f66296b0d550a +size 1280828 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample60-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample60-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1fc2e7cd4d3aaf2c0853a674361e7ca54924364d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample60-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:826b70ec1ba8178262a8c12f72ffbb9f019c40fb35cea435640e7259594076b0 +size 2043004 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample600-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample600-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..bc1e5ca5a36af3286d825ba8c4218b04db47634a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample600-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a2d82220dc5f191349a9f88123e7a79cf3ee9e5de1a16a7be660ab5c5b2eff0 +size 1268101 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample622-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample622-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e501f04bf01c985c7e42d3dc89cfe42957fd1c4e --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample622-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a47b266cc20314c6e275f15a8e5efbebb7361faa9eb3fadcb4a8aab76a159995 +size 1302223 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample65-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample65-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..95fd9f12419e0a1e4143a1c77ea9f106f66ff97a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample65-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1425de5990a23a86f76bea54686e0a3aa0eb5c062df234a57b8fefb4ddab0b56 +size 2001036 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample656-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample656-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7517ab6d3131d36305afe5a49df4f169948fb33c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample656-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cef694e75e2be33f74c2f42f8f15ad00be01686c03cc2115a023f637e44bbd9b +size 1218120 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample66-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample66-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..9bf5f77507dcc4d6934026ae0928e9449099fb7a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample66-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57b986f80a977b24afd00a4fbebbde4ad66c59deba340101b0007e500b592fca +size 2414375 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample663-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample663-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..62555890178fc40b0de210aa7e603619cd063b49 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample663-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72f472a3197a12f544f92d230b92e2419b53e980431e204c61fc933640b618b5 +size 1202481 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample68-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample68-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c52d1a628d9484540cc045a65f5349a8c0d8cedf --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample68-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:592cfcc2dfb6a7bd801a0ab10051fb4a962252b97f61607cd2354392f56c5132 +size 2174011 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample7-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample7-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3defa167053062bd006706fd54c007d78bd66f39 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample7-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5051a79ef4d02ebed6cada17624606b160707752c602d3ed8dea3318b43248d8 +size 2773882 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample736-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample736-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..6530a8f1bbf8734ac0f6836eb288262e2e683893 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample736-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c653f1ceb1a4bfc9eba3f34fc6bf5f984dbbb0e1b229b994cffd30af435c4a4 +size 1340948 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample74-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample74-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0b143c701c61e0bddb6f9eb8f9e4459961d8b63a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample74-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:914303b2f00f738ef0d61ff980b0eec1cb043e1606213a9debddef8c425e26f2 +size 2175356 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample75-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample75-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..88ec1c591edaa7c3fb90ec89b6d4acb9cfc7699c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample75-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2b55229cb56cda98233c9619bed9656a3e38d5887a1ff2781a5a7161ed4da61 +size 2524996 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample777-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample777-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..9647ca83dddb0347e80b5629ef05a5973f896e06 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample777-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6ab3df856335b199a5ec4e689a52ab3aefd2cbd6e2fbdb38d1e11357e94ab0b +size 1316267 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample778-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample778-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f289aebf899b464a716d51f0a04ed5f6a69c0489 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample778-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19e80f1e4bd2cbc0d5e84c2b40030b11ead0569de65e9730e2fe25ea1b389b8e +size 1099208 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample807-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample807-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..fee261813c11816c1f74baee4c8bcb0f73454900 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample807-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0575fe7fd23bcb46590edd3339d3cb5b940f14252bba5a3f7dddc6d7f7029d13 +size 1066753 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample855-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample855-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1d887ec29a42cc989a583655211332de298b7ead --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample855-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f77192398b532e220272bdaa13a8f17260c8aeb8d33f27d0cb0eade30493399a +size 1173022 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample859-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample859-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..51722d1327c42a6a62a9941a62bb75dd777e2dc5 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample859-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc681fa9b815eb0267b4a2e351595081ea635ff15b39624f74074e20a92a9fd7 +size 1297962 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample86-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample86-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..01567d9ec7860562d37ba4484bea72d3c3ffbdc8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample86-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f282b557dd0cede7ebfacb65c157dcbaefa7c11c08fe9d89a1e34fa03a644642 +size 2145584 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample91-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample91-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f6df913b679ecf58e9643f20ca5f17be1aabc3cd --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample91-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc66aa152326a356a37fc908d2153e6eb6641e38cb4193a7976da5d6448a6152 +size 2000288 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample92-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample92-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..003aef6e0f4ad94973451e979d3cf2a22197bbbf --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample92-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ef0d168eaa985c5ca1b6271daed3ceba2ae83f04d13542cc05931f7358846bf +size 1946297 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample925-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample925-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b99373b56b2647678e737bed324d0435380371d7 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample925-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81ff7fe79bb88784eed674270505d3de909af6f50fbd3ea5b8d400be754e6829 +size 1164295 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample95-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample95-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0ea22d501fc30fffdafb664896e8adcc64a9715e --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample95-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9278fc0d3d6fcf145b7c47d82c83846c7520868a3d7a9b1c68538df63491a5b +size 2058531 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample967-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample967-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..74bb083833e281a65f0bcf81743c4d1edcb90e23 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample967-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4fcb269f5abecfe5c7aba1409e29bd0b15f8703ffe2c83a7e94f958ce30ae936 +size 1058294 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample969-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample969-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e0829da5a3511284dc0e1d14873c0fe2910f2299 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample969-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9dcad0ad78677f5f4342f8bc719d5f32994096ddf77133a82774ec5ec37eded5 +size 1083902 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample97-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample97-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..63a55f28bfd79be7e92e63415e7d4dba2a265303 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_gsm8k/sample97-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d8802f6d27883bdb8c48917293ce773561a1409c8c041c7abcbc12c8a8b14aaa +size 2011337 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/dtufc_hyperprior-featurecoding_qwen_individual.log b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/dtufc_hyperprior-featurecoding_qwen_individual.log index 0c6435b0a21e2e0a466bd1fe7d50e631c730d7d2..1f189ef64231069a0d0660f8705af74403028f34 100644 --- a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/dtufc_hyperprior-featurecoding_qwen_individual.log +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/dtufc_hyperprior-featurecoding_qwen_individual.log @@ -1,16958 +1,3 @@ -Experiment: dtufc_hyperprior-featurecoding_qwen_individual -Log file: output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/dtufc_hyperprior-featurecoding_qwen_individual.log -DTUFCCodecConfig: - arch: hyperprior-featurecoding - handler: qwen - checkpoint: codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar - transform_type: kmeans - transform_mapping:featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json - bit_depth: 8 - device: cuda:0 -Loading checkpoint: codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar -Checkpoint epoch: 599 -Loaded hyperprior-featurecoding (1-channel) on cuda:0 -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.0.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.0.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.1.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.1.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.2.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.2.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.3.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.3.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json -Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json: torch.Size([256]) -Loaded per-key quantization points for key 'layer.4.output' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json -Loaded per-key mappings: model=qwen - Keys: ['layer.0.k_cache', 'layer.0.v_cache', 'layer.1.k_cache', 'layer.1.v_cache', 'layer.2.k_cache', 'layer.2.v_cache', 'layer.3.k_cache', 'layer.3.v_cache', 'layer.4.k_cache', 'layer.4.v_cache', 'layer.4.output'] ----------------- ----------------------------------------------------------------------------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding -Checkpoint codec_weights/hyperprior_hybrid/bmshj2018-hyperprior_lambda0.01_epochs600_lr0.0001_bs360_patch256-256_checkpoint_best.pth.tar -Transform type kmeans -Transform config featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json -Input ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag -Output output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag ----------------- ----------------------------------------------------------------------------------------------------------------------------- -Files found: 100 ----------------------------------------------------------------------- - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample0-layer4-item1.zst (1/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample0-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 371, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 371, 128) -Output shape: (1, 371, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.0.v_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.1.k_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.1.v_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.2.k_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.2.v_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.3.k_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.3.v_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.4.k_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.4.v_cache: torch.Size([1, 8, 371, 128]) -> torch.Size([1, 1, 371, 1024]) - layer.4.output: torch.Size([1, 371, 4096]) -> torch.Size([1, 1, 371, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,760B, BPFP=0.6899 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 86,300B, BPFP=1.8173 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,600B, BPFP=1.2761 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 92,232B, BPFP=1.9422 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,756B, BPFP=1.4479 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 94,020B, BPFP=1.9799 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,772B, BPFP=1.4903 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 92,624B, BPFP=1.9505 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 62,116B, BPFP=1.3080 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 94,612B, BPFP=1.9923 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 268,148B, BPFP=1.4117 -⌛️ [2/4] FRONTEND: Frontend time: 0.661s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 371, 128]) - layer.0.v_cache: torch.Size([1, 8, 371, 128]) - layer.1.k_cache: torch.Size([1, 8, 371, 128]) - layer.1.v_cache: torch.Size([1, 8, 371, 128]) - layer.2.k_cache: torch.Size([1, 8, 371, 128]) - layer.2.v_cache: torch.Size([1, 8, 371, 128]) - layer.3.k_cache: torch.Size([1, 8, 371, 128]) - layer.3.v_cache: torch.Size([1, 8, 371, 128]) - layer.4.k_cache: torch.Size([1, 8, 371, 128]) - layer.4.v_cache: torch.Size([1, 8, 371, 128]) - layer.4.output: torch.Size([1, 371, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.849s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 371, 128]) - layer.0.v_cache: torch.Size([1, 8, 371, 128]) - layer.1.k_cache: torch.Size([1, 8, 371, 128]) - layer.1.v_cache: torch.Size([1, 8, 371, 128]) - layer.2.k_cache: torch.Size([1, 8, 371, 128]) - layer.2.v_cache: torch.Size([1, 8, 371, 128]) - layer.3.k_cache: torch.Size([1, 8, 371, 128]) - layer.3.v_cache: torch.Size([1, 8, 371, 128]) - layer.4.k_cache: torch.Size([1, 8, 371, 128]) - layer.4.v_cache: torch.Size([1, 8, 371, 128]) - layer.4.output: torch.Size([1, 371, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02382561 6.66727186 - layer.0.v_cache 0.00000026 0.00022253 - layer.1.k_cache 0.00289382 0.61497247 - layer.1.v_cache 0.00000075 0.00080611 - layer.2.k_cache 0.00115144 0.39387915 - layer.2.v_cache 0.00000114 0.00116830 - layer.3.k_cache 0.00133317 0.43466413 - layer.3.v_cache 0.00000213 0.00188232 - layer.4.k_cache 0.00354181 0.76364621 - layer.4.v_cache 0.00000324 0.00318926 - layer.4.output 0.00015639 0.05413258 - ------------------------------------------------------------------------------------- - TOTAL 0.00238421 0.64987376 - (elements=5,318,656) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5318656 -Total Bytes 1022940 -BPFP 1.5386 bits/point -EBPFP 3.0773 equivalent bits/point -MSE 0.649874 ----------------------- -------------------------------------------------------- -Time: 1.530s Load: 0.019s, Pack+Encode: 0.661s, Decode+Unpack: 0.849s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 371, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 371, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6499 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample0-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample0-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample1-layer4-item1.zst (2/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample1-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 344, 128) -Output shape: (1, 344, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.output: torch.Size([1, 344, 4096]) -> torch.Size([1, 1, 344, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,960B, BPFP=0.7258 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,816B, BPFP=1.8581 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,504B, BPFP=1.3060 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,696B, BPFP=1.9916 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,112B, BPFP=1.4787 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,652B, BPFP=2.0361 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,896B, BPFP=1.5193 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,212B, BPFP=2.0034 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,556B, BPFP=1.3299 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,052B, BPFP=2.0451 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 256,236B, BPFP=1.4548 -⌛️ [2/4] FRONTEND: Frontend time: 0.427s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.721s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02409526 6.91969690 - layer.0.v_cache 0.00000026 0.00022533 - layer.1.k_cache 0.00287302 0.59615086 - layer.1.v_cache 0.00000078 0.00082355 - layer.2.k_cache 0.00123767 0.38879457 - layer.2.v_cache 0.00000117 0.00123389 - layer.3.k_cache 0.00130179 0.42716820 - layer.3.v_cache 0.00000226 0.00197517 - layer.4.k_cache 0.00355144 0.74402716 - layer.4.v_cache 0.00000325 0.00334561 - layer.4.output 0.00015063 0.05544508 - ------------------------------------------------------------------------------------- - TOTAL 0.00240496 0.66465868 - (elements=4,931,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4931584 -Total Bytes 973692 -BPFP 1.5795 bits/point -EBPFP 3.1590 equivalent bits/point -MSE 0.664659 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.019s, Pack+Encode: 0.427s, Decode+Unpack: 0.721s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6647 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample1-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample1-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample10-layer4-item1.zst (3/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample10-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 377, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.026s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 377, 128) -Output shape: (1, 377, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.0.v_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.1.k_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.1.v_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.2.k_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.2.v_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.3.k_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.3.v_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.4.k_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.4.v_cache: torch.Size([1, 8, 377, 128]) -> torch.Size([1, 1, 377, 1024]) - layer.4.output: torch.Size([1, 377, 4096]) -> torch.Size([1, 1, 377, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 33,972B, BPFP=0.7040 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 87,704B, BPFP=1.8175 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 62,136B, BPFP=1.2876 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 93,988B, BPFP=1.9477 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 70,232B, BPFP=1.4554 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 95,412B, BPFP=1.9772 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 71,792B, BPFP=1.4877 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 93,532B, BPFP=1.9382 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 63,160B, BPFP=1.3089 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 96,092B, BPFP=1.9913 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 274,088B, BPFP=1.4200 -⌛️ [2/4] FRONTEND: Frontend time: 0.450s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 377, 128]) - layer.0.v_cache: torch.Size([1, 8, 377, 128]) - layer.1.k_cache: torch.Size([1, 8, 377, 128]) - layer.1.v_cache: torch.Size([1, 8, 377, 128]) - layer.2.k_cache: torch.Size([1, 8, 377, 128]) - layer.2.v_cache: torch.Size([1, 8, 377, 128]) - layer.3.k_cache: torch.Size([1, 8, 377, 128]) - layer.3.v_cache: torch.Size([1, 8, 377, 128]) - layer.4.k_cache: torch.Size([1, 8, 377, 128]) - layer.4.v_cache: torch.Size([1, 8, 377, 128]) - layer.4.output: torch.Size([1, 377, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.710s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 377, 128]) - layer.0.v_cache: torch.Size([1, 8, 377, 128]) - layer.1.k_cache: torch.Size([1, 8, 377, 128]) - layer.1.v_cache: torch.Size([1, 8, 377, 128]) - layer.2.k_cache: torch.Size([1, 8, 377, 128]) - layer.2.v_cache: torch.Size([1, 8, 377, 128]) - layer.3.k_cache: torch.Size([1, 8, 377, 128]) - layer.3.v_cache: torch.Size([1, 8, 377, 128]) - layer.4.k_cache: torch.Size([1, 8, 377, 128]) - layer.4.v_cache: torch.Size([1, 8, 377, 128]) - layer.4.output: torch.Size([1, 377, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02400962 6.61519358 - layer.0.v_cache 0.00000027 0.00023091 - layer.1.k_cache 0.00291640 0.60843519 - layer.1.v_cache 0.00000080 0.00086127 - layer.2.k_cache 0.00116459 0.39109980 - layer.2.v_cache 0.00000114 0.00119852 - layer.3.k_cache 0.00131780 0.42198201 - layer.3.v_cache 0.00000210 0.00190877 - layer.4.k_cache 0.00355012 0.75056445 - layer.4.v_cache 0.00000332 0.00340337 - layer.4.output 0.00014433 0.05150712 - ------------------------------------------------------------------------------------- - TOTAL 0.00239596 0.64292188 - (elements=5,404,672) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5404672 -Total Bytes 1042108 -BPFP 1.5425 bits/point -EBPFP 3.0851 equivalent bits/point -MSE 0.642922 ----------------------- -------------------------------------------------------- -Time: 1.187s Load: 0.026s, Pack+Encode: 0.450s, Decode+Unpack: 0.710s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 377, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 377, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6429 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample10-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample10-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample100-layer4-item1.zst (4/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample100-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 315, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.020s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 315, 128) -Output shape: (1, 315, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.0.v_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.1.k_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.1.v_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.2.k_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.2.v_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.3.k_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.3.v_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.4.k_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.4.v_cache: torch.Size([1, 8, 315, 128]) -> torch.Size([1, 1, 315, 1024]) - layer.4.output: torch.Size([1, 315, 4096]) -> torch.Size([1, 1, 315, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,136B, BPFP=0.7226 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,496B, BPFP=1.8228 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,076B, BPFP=1.2916 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,380B, BPFP=1.9439 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,808B, BPFP=1.4585 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,516B, BPFP=1.9721 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,160B, BPFP=1.4921 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,360B, BPFP=1.9435 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,196B, BPFP=1.3193 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 80,040B, BPFP=1.9851 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 231,096B, BPFP=1.4329 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 315, 128]) - layer.0.v_cache: torch.Size([1, 8, 315, 128]) - layer.1.k_cache: torch.Size([1, 8, 315, 128]) - layer.1.v_cache: torch.Size([1, 8, 315, 128]) - layer.2.k_cache: torch.Size([1, 8, 315, 128]) - layer.2.v_cache: torch.Size([1, 8, 315, 128]) - layer.3.k_cache: torch.Size([1, 8, 315, 128]) - layer.3.v_cache: torch.Size([1, 8, 315, 128]) - layer.4.k_cache: torch.Size([1, 8, 315, 128]) - layer.4.v_cache: torch.Size([1, 8, 315, 128]) - layer.4.output: torch.Size([1, 315, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.596s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 315, 128]) - layer.0.v_cache: torch.Size([1, 8, 315, 128]) - layer.1.k_cache: torch.Size([1, 8, 315, 128]) - layer.1.v_cache: torch.Size([1, 8, 315, 128]) - layer.2.k_cache: torch.Size([1, 8, 315, 128]) - layer.2.v_cache: torch.Size([1, 8, 315, 128]) - layer.3.k_cache: torch.Size([1, 8, 315, 128]) - layer.3.v_cache: torch.Size([1, 8, 315, 128]) - layer.4.k_cache: torch.Size([1, 8, 315, 128]) - layer.4.v_cache: torch.Size([1, 8, 315, 128]) - layer.4.output: torch.Size([1, 315, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02468849 6.65798301 - layer.0.v_cache 0.00000027 0.00023166 - layer.1.k_cache 0.00284817 0.60716892 - layer.1.v_cache 0.00000082 0.00085317 - layer.2.k_cache 0.00122538 0.38983317 - layer.2.v_cache 0.00000116 0.00119732 - layer.3.k_cache 0.00131111 0.42748224 - layer.3.v_cache 0.00000216 0.00195930 - layer.4.k_cache 0.00356899 0.75562555 - layer.4.v_cache 0.00000330 0.00333117 - layer.4.output 0.00013845 0.05498815 - ------------------------------------------------------------------------------------- - TOTAL 0.00244312 0.64754415 - (elements=4,515,840) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4515840 -Total Bytes 874264 -BPFP 1.5488 bits/point -EBPFP 3.0976 equivalent bits/point -MSE 0.647544 ----------------------- -------------------------------------------------------- -Time: 1.045s Load: 0.020s, Pack+Encode: 0.429s, Decode+Unpack: 0.596s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 315, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 315, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6475 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample100-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample100-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample102-layer4-item1.zst (5/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample102-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 362, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 362, 128) -Output shape: (1, 362, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.output: torch.Size([1, 362, 4096]) -> torch.Size([1, 1, 362, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,584B, BPFP=0.6816 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 85,412B, BPFP=1.8433 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,208B, BPFP=1.2994 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 92,308B, BPFP=1.9921 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,252B, BPFP=1.4730 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 93,632B, BPFP=2.0207 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,176B, BPFP=1.5145 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 92,044B, BPFP=1.9864 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,324B, BPFP=1.3235 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 94,180B, BPFP=2.0325 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 260,916B, BPFP=1.4077 -⌛️ [2/4] FRONTEND: Frontend time: 0.423s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.output: torch.Size([1, 362, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.697s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.output: torch.Size([1, 362, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02398593 7.05788224 - layer.0.v_cache 0.00000027 0.00023763 - layer.1.k_cache 0.00294127 0.63709891 - layer.1.v_cache 0.00000079 0.00085997 - layer.2.k_cache 0.00116280 0.39235835 - layer.2.v_cache 0.00000114 0.00120453 - layer.3.k_cache 0.00131002 0.42396798 - layer.3.v_cache 0.00000215 0.00195116 - layer.4.k_cache 0.00366934 0.78028043 - layer.4.v_cache 0.00000328 0.00336954 - layer.4.output 0.00013411 0.05014663 - ------------------------------------------------------------------------------------- - TOTAL 0.00240096 0.67855695 - (elements=5,189,632) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5189632 -Total Bytes 1010036 -BPFP 1.5570 bits/point -EBPFP 3.1140 equivalent bits/point -MSE 0.678557 ----------------------- -------------------------------------------------------- -Time: 1.139s Load: 0.019s, Pack+Encode: 0.423s, Decode+Unpack: 0.697s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 362, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6786 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample102-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample102-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample105-layer4-item1.zst (6/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample105-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 336, 128) -Output shape: (1, 336, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.output: torch.Size([1, 336, 4096]) -> torch.Size([1, 1, 336, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,688B, BPFP=0.7135 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,828B, BPFP=1.8561 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,028B, BPFP=1.3027 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,772B, BPFP=1.9943 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,032B, BPFP=1.4656 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,676B, BPFP=2.0386 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,632B, BPFP=1.5260 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,356B, BPFP=2.0079 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,820B, BPFP=1.3211 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,008B, BPFP=2.0463 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 245,872B, BPFP=1.4292 -⌛️ [2/4] FRONTEND: Frontend time: 0.412s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.696s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02466884 7.06196304 - layer.0.v_cache 0.00000026 0.00022800 - layer.1.k_cache 0.00293366 0.64410092 - layer.1.v_cache 0.00000078 0.00084075 - layer.2.k_cache 0.00117059 0.38787647 - layer.2.v_cache 0.00000113 0.00121264 - layer.3.k_cache 0.00132874 0.43136306 - layer.3.v_cache 0.00000215 0.00196611 - layer.4.k_cache 0.00360252 0.76547623 - layer.4.v_cache 0.00000314 0.00318044 - layer.4.output 0.00014135 0.05365221 - ------------------------------------------------------------------------------------- - TOTAL 0.00244837 0.67948689 - (elements=4,816,896) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4816896 -Total Bytes 945712 -BPFP 1.5707 bits/point -EBPFP 3.1413 equivalent bits/point -MSE 0.679487 ----------------------- -------------------------------------------------------- -Time: 1.127s Load: 0.018s, Pack+Encode: 0.412s, Decode+Unpack: 0.696s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6795 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample105-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample105-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample108-layer4-item1.zst (7/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample108-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 317, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 317, 128) -Output shape: (1, 317, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.0.v_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.1.k_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.1.v_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.2.k_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.2.v_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.3.k_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.3.v_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.4.k_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.4.v_cache: torch.Size([1, 8, 317, 128]) -> torch.Size([1, 1, 317, 1024]) - layer.4.output: torch.Size([1, 317, 4096]) -> torch.Size([1, 1, 317, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,716B, BPFP=0.7077 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,724B, BPFP=1.7923 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,280B, BPFP=1.2884 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,084B, BPFP=1.9244 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,872B, BPFP=1.4509 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,220B, BPFP=1.9524 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,244B, BPFP=1.4847 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,100B, BPFP=1.9248 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,536B, BPFP=1.3194 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,796B, BPFP=1.9666 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 230,880B, BPFP=1.4225 -⌛️ [2/4] FRONTEND: Frontend time: 0.370s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 317, 128]) - layer.0.v_cache: torch.Size([1, 8, 317, 128]) - layer.1.k_cache: torch.Size([1, 8, 317, 128]) - layer.1.v_cache: torch.Size([1, 8, 317, 128]) - layer.2.k_cache: torch.Size([1, 8, 317, 128]) - layer.2.v_cache: torch.Size([1, 8, 317, 128]) - layer.3.k_cache: torch.Size([1, 8, 317, 128]) - layer.3.v_cache: torch.Size([1, 8, 317, 128]) - layer.4.k_cache: torch.Size([1, 8, 317, 128]) - layer.4.v_cache: torch.Size([1, 8, 317, 128]) - layer.4.output: torch.Size([1, 317, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.596s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 317, 128]) - layer.0.v_cache: torch.Size([1, 8, 317, 128]) - layer.1.k_cache: torch.Size([1, 8, 317, 128]) - layer.1.v_cache: torch.Size([1, 8, 317, 128]) - layer.2.k_cache: torch.Size([1, 8, 317, 128]) - layer.2.v_cache: torch.Size([1, 8, 317, 128]) - layer.3.k_cache: torch.Size([1, 8, 317, 128]) - layer.3.v_cache: torch.Size([1, 8, 317, 128]) - layer.4.k_cache: torch.Size([1, 8, 317, 128]) - layer.4.v_cache: torch.Size([1, 8, 317, 128]) - layer.4.output: torch.Size([1, 317, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02499472 6.66056366 - layer.0.v_cache 0.00000027 0.00023465 - layer.1.k_cache 0.00294942 0.62116364 - layer.1.v_cache 0.00000081 0.00087078 - layer.2.k_cache 0.00116222 0.39341563 - layer.2.v_cache 0.00000115 0.00124362 - layer.3.k_cache 0.00131165 0.43894376 - layer.3.v_cache 0.00000215 0.00202564 - layer.4.k_cache 0.00349495 0.76082922 - layer.4.v_cache 0.00000332 0.00344669 - layer.4.output 0.00014092 0.05730051 - ------------------------------------------------------------------------------------- - TOTAL 0.00246317 0.65085281 - (elements=4,544,512) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4544512 -Total Bytes 872452 -BPFP 1.5358 bits/point -EBPFP 3.0717 equivalent bits/point -MSE 0.650853 ----------------------- -------------------------------------------------------- -Time: 0.982s Load: 0.017s, Pack+Encode: 0.370s, Decode+Unpack: 0.596s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 317, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 317, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6509 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample108-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample108-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample11-layer4-item1.zst (8/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample11-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 353, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.023s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 353, 128) -Output shape: (1, 353, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.output: torch.Size([1, 353, 4096]) -> torch.Size([1, 1, 353, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,460B, BPFP=0.6963 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,512B, BPFP=1.8261 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,100B, BPFP=1.2859 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 89,068B, BPFP=1.9712 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,760B, BPFP=1.4554 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,608B, BPFP=2.0053 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,756B, BPFP=1.4996 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,868B, BPFP=1.9668 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,148B, BPFP=1.3090 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 91,056B, BPFP=2.0152 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 250,648B, BPFP=1.3868 -⌛️ [2/4] FRONTEND: Frontend time: 0.432s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.output: torch.Size([1, 353, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.696s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.output: torch.Size([1, 353, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02498756 6.80341700 - layer.0.v_cache 0.00000027 0.00023479 - layer.1.k_cache 0.00288490 0.62682297 - layer.1.v_cache 0.00000079 0.00089020 - layer.2.k_cache 0.00118233 0.39887153 - layer.2.v_cache 0.00000115 0.00125327 - layer.3.k_cache 0.00132883 0.44330854 - layer.3.v_cache 0.00000217 0.00193058 - layer.4.k_cache 0.00368366 0.77350173 - layer.4.v_cache 0.00000317 0.00330873 - layer.4.output 0.00013605 0.05182723 - ------------------------------------------------------------------------------------- - TOTAL 0.00247279 0.66148916 - (elements=5,060,608) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5060608 -Total Bytes 974984 -BPFP 1.5413 bits/point -EBPFP 3.0826 equivalent bits/point -MSE 0.661489 ----------------------- -------------------------------------------------------- -Time: 1.151s Load: 0.023s, Pack+Encode: 0.432s, Decode+Unpack: 0.696s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 353, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6615 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample11-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample11-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample112-layer4-item1.zst (9/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample112-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 318, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 318, 128) -Output shape: (1, 318, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.output: torch.Size([1, 318, 4096]) -> torch.Size([1, 1, 318, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,528B, BPFP=0.7254 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,776B, BPFP=1.7879 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,232B, BPFP=1.2832 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 76,676B, BPFP=1.8837 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,428B, BPFP=1.4354 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 78,316B, BPFP=1.9240 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,504B, BPFP=1.4864 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,492B, BPFP=1.9038 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,560B, BPFP=1.3158 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,384B, BPFP=1.9503 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 234,720B, BPFP=1.4416 -⌛️ [2/4] FRONTEND: Frontend time: 0.378s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.output: torch.Size([1, 318, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.593s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.output: torch.Size([1, 318, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02461332 6.43196998 - layer.0.v_cache 0.00000025 0.00021872 - layer.1.k_cache 0.00290030 0.63886909 - layer.1.v_cache 0.00000077 0.00076085 - layer.2.k_cache 0.00116467 0.39368472 - layer.2.v_cache 0.00000108 0.00111194 - layer.3.k_cache 0.00132360 0.44378499 - layer.3.v_cache 0.00000220 0.00186514 - layer.4.k_cache 0.00367546 0.79008541 - layer.4.v_cache 0.00000314 0.00320942 - layer.4.output 0.00019099 0.06342408 - ------------------------------------------------------------------------------------- - TOTAL 0.00246062 0.63994690 - (elements=4,558,848) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4558848 -Total Bytes 873616 -BPFP 1.5330 bits/point -EBPFP 3.0661 equivalent bits/point -MSE 0.639947 ----------------------- -------------------------------------------------------- -Time: 0.988s Load: 0.017s, Pack+Encode: 0.378s, Decode+Unpack: 0.593s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 318, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6399 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample112-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample112-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample114-layer4-item1.zst (10/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample114-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 330, 128) -Output shape: (1, 330, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.output: torch.Size([1, 330, 4096]) -> torch.Size([1, 1, 330, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,640B, BPFP=0.7254 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,152B, BPFP=1.8739 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,268B, BPFP=1.3084 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,292B, BPFP=2.0192 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,800B, BPFP=1.4867 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,076B, BPFP=2.0615 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,576B, BPFP=1.5288 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,512B, BPFP=2.0244 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,500B, BPFP=1.3376 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,632B, BPFP=2.0746 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,252B, BPFP=1.4693 -⌛️ [2/4] FRONTEND: Frontend time: 0.441s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.710s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02377391 7.00123328 - layer.0.v_cache 0.00000027 0.00023593 - layer.1.k_cache 0.00289258 0.62951415 - layer.1.v_cache 0.00000083 0.00085029 - layer.2.k_cache 0.00117497 0.39299617 - layer.2.v_cache 0.00000126 0.00123513 - layer.3.k_cache 0.00131231 0.42370800 - layer.3.v_cache 0.00000230 0.00200699 - layer.4.k_cache 0.00357793 0.77173189 - layer.4.v_cache 0.00000347 0.00342087 - layer.4.output 0.00014070 0.05641136 - ------------------------------------------------------------------------------------- - TOTAL 0.00237876 0.67518415 - (elements=4,730,880) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4730880 -Total Bytes 942700 -BPFP 1.5941 bits/point -EBPFP 3.1882 equivalent bits/point -MSE 0.675184 ----------------------- -------------------------------------------------------- -Time: 1.168s Load: 0.018s, Pack+Encode: 0.441s, Decode+Unpack: 0.710s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6752 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample114-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample114-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample119-layer4-item1.zst (11/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample119-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 357, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 357, 128) -Output shape: (1, 357, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.0.v_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.1.k_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.1.v_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.2.k_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.2.v_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.3.k_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.3.v_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.4.k_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.4.v_cache: torch.Size([1, 8, 357, 128]) -> torch.Size([1, 1, 357, 1024]) - layer.4.output: torch.Size([1, 357, 4096]) -> torch.Size([1, 1, 357, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 33,056B, BPFP=0.7234 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 85,428B, BPFP=1.8695 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 59,448B, BPFP=1.3009 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 90,904B, BPFP=1.9893 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 67,644B, BPFP=1.4803 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 92,612B, BPFP=2.0267 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 69,620B, BPFP=1.5235 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 91,420B, BPFP=2.0006 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 60,624B, BPFP=1.3267 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 93,468B, BPFP=2.0454 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 270,492B, BPFP=1.4798 -⌛️ [2/4] FRONTEND: Frontend time: 0.419s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 357, 128]) - layer.0.v_cache: torch.Size([1, 8, 357, 128]) - layer.1.k_cache: torch.Size([1, 8, 357, 128]) - layer.1.v_cache: torch.Size([1, 8, 357, 128]) - layer.2.k_cache: torch.Size([1, 8, 357, 128]) - layer.2.v_cache: torch.Size([1, 8, 357, 128]) - layer.3.k_cache: torch.Size([1, 8, 357, 128]) - layer.3.v_cache: torch.Size([1, 8, 357, 128]) - layer.4.k_cache: torch.Size([1, 8, 357, 128]) - layer.4.v_cache: torch.Size([1, 8, 357, 128]) - layer.4.output: torch.Size([1, 357, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.710s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 357, 128]) - layer.0.v_cache: torch.Size([1, 8, 357, 128]) - layer.1.k_cache: torch.Size([1, 8, 357, 128]) - layer.1.v_cache: torch.Size([1, 8, 357, 128]) - layer.2.k_cache: torch.Size([1, 8, 357, 128]) - layer.2.v_cache: torch.Size([1, 8, 357, 128]) - layer.3.k_cache: torch.Size([1, 8, 357, 128]) - layer.3.v_cache: torch.Size([1, 8, 357, 128]) - layer.4.k_cache: torch.Size([1, 8, 357, 128]) - layer.4.v_cache: torch.Size([1, 8, 357, 128]) - layer.4.output: torch.Size([1, 357, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02468803 7.23039352 - layer.0.v_cache 0.00000027 0.00023325 - layer.1.k_cache 0.00290374 0.64391419 - layer.1.v_cache 0.00000082 0.00083659 - layer.2.k_cache 0.00112985 0.38967348 - layer.2.v_cache 0.00000115 0.00115126 - layer.3.k_cache 0.00131122 0.43220110 - layer.3.v_cache 0.00000220 0.00194585 - layer.4.k_cache 0.00345488 0.76421093 - layer.4.v_cache 0.00000325 0.00328564 - layer.4.output 0.00016267 0.06167226 - ------------------------------------------------------------------------------------- - TOTAL 0.00243901 0.69389535 - (elements=5,117,952) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5117952 -Total Bytes 1014716 -BPFP 1.5861 bits/point -EBPFP 3.1723 equivalent bits/point -MSE 0.693895 ----------------------- -------------------------------------------------------- -Time: 1.147s Load: 0.019s, Pack+Encode: 0.419s, Decode+Unpack: 0.710s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 357, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 357, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6939 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample119-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample119-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample12-layer4-item1.zst (12/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample12-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 358, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 358, 128) -Output shape: (1, 358, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.0.v_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.1.k_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.1.v_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.2.k_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.2.v_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.3.k_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.3.v_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.4.k_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.4.v_cache: torch.Size([1, 8, 358, 128]) -> torch.Size([1, 1, 358, 1024]) - layer.4.output: torch.Size([1, 358, 4096]) -> torch.Size([1, 1, 358, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,712B, BPFP=0.7139 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 84,764B, BPFP=1.8498 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 59,444B, BPFP=1.2972 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 91,004B, BPFP=1.9859 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 67,452B, BPFP=1.4720 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 92,792B, BPFP=2.0250 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 69,176B, BPFP=1.5096 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 91,336B, BPFP=1.9932 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 60,796B, BPFP=1.3267 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 93,744B, BPFP=2.0457 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 264,928B, BPFP=1.4454 -⌛️ [2/4] FRONTEND: Frontend time: 0.441s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 358, 128]) - layer.0.v_cache: torch.Size([1, 8, 358, 128]) - layer.1.k_cache: torch.Size([1, 8, 358, 128]) - layer.1.v_cache: torch.Size([1, 8, 358, 128]) - layer.2.k_cache: torch.Size([1, 8, 358, 128]) - layer.2.v_cache: torch.Size([1, 8, 358, 128]) - layer.3.k_cache: torch.Size([1, 8, 358, 128]) - layer.3.v_cache: torch.Size([1, 8, 358, 128]) - layer.4.k_cache: torch.Size([1, 8, 358, 128]) - layer.4.v_cache: torch.Size([1, 8, 358, 128]) - layer.4.output: torch.Size([1, 358, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.706s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 358, 128]) - layer.0.v_cache: torch.Size([1, 8, 358, 128]) - layer.1.k_cache: torch.Size([1, 8, 358, 128]) - layer.1.v_cache: torch.Size([1, 8, 358, 128]) - layer.2.k_cache: torch.Size([1, 8, 358, 128]) - layer.2.v_cache: torch.Size([1, 8, 358, 128]) - layer.3.k_cache: torch.Size([1, 8, 358, 128]) - layer.3.v_cache: torch.Size([1, 8, 358, 128]) - layer.4.k_cache: torch.Size([1, 8, 358, 128]) - layer.4.v_cache: torch.Size([1, 8, 358, 128]) - layer.4.output: torch.Size([1, 358, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02418174 7.02813823 - layer.0.v_cache 0.00000026 0.00023389 - layer.1.k_cache 0.00288075 0.63533135 - layer.1.v_cache 0.00000077 0.00084294 - layer.2.k_cache 0.00118336 0.38929540 - layer.2.v_cache 0.00000117 0.00121996 - layer.3.k_cache 0.00130988 0.43330038 - layer.3.v_cache 0.00000211 0.00193445 - layer.4.k_cache 0.00353008 0.75301323 - layer.4.v_cache 0.00000317 0.00337383 - layer.4.output 0.00013834 0.05213491 - ------------------------------------------------------------------------------------- - TOTAL 0.00240333 0.67537309 - (elements=5,132,288) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5132288 -Total Bytes 1008148 -BPFP 1.5715 bits/point -EBPFP 3.1429 equivalent bits/point -MSE 0.675373 ----------------------- -------------------------------------------------------- -Time: 1.165s Load: 0.018s, Pack+Encode: 0.441s, Decode+Unpack: 0.706s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 358, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 358, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6754 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample12-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample12-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample13-layer4-item1.zst (13/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample13-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 336, 128) -Output shape: (1, 336, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.output: torch.Size([1, 336, 4096]) -> torch.Size([1, 1, 336, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,776B, BPFP=0.7156 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,620B, BPFP=1.8513 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,816B, BPFP=1.2978 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,088B, BPFP=2.0017 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,376B, BPFP=1.4736 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,916B, BPFP=2.0442 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,756B, BPFP=1.5289 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,468B, BPFP=2.0105 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,968B, BPFP=1.3246 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,468B, BPFP=2.0570 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 251,060B, BPFP=1.4594 -⌛️ [2/4] FRONTEND: Frontend time: 0.419s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02407829 6.94693284 - layer.0.v_cache 0.00000026 0.00022575 - layer.1.k_cache 0.00285290 0.65468125 - layer.1.v_cache 0.00000077 0.00085141 - layer.2.k_cache 0.00116461 0.39580254 - layer.2.v_cache 0.00000115 0.00126206 - layer.3.k_cache 0.00131253 0.43582984 - layer.3.v_cache 0.00000220 0.00199418 - layer.4.k_cache 0.00369396 0.78648640 - layer.4.v_cache 0.00000328 0.00343792 - layer.4.output 0.00015589 0.06668643 - ------------------------------------------------------------------------------------- - TOTAL 0.00240954 0.67816071 - (elements=4,816,896) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4816896 -Total Bytes 952312 -BPFP 1.5816 bits/point -EBPFP 3.1632 equivalent bits/point -MSE 0.678161 ----------------------- -------------------------------------------------------- -Time: 1.142s Load: 0.017s, Pack+Encode: 0.419s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6782 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample13-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample13-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample132-layer4-item1.zst (14/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample132-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 346, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 346, 128) -Output shape: (1, 346, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.output: torch.Size([1, 346, 4096]) -> torch.Size([1, 1, 346, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,012B, BPFP=0.7228 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,756B, BPFP=1.8460 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,148B, BPFP=1.2904 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,376B, BPFP=1.9729 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,892B, BPFP=1.4652 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,256B, BPFP=2.0154 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,920B, BPFP=1.5110 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,128B, BPFP=1.9899 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,296B, BPFP=1.3163 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,156B, BPFP=2.0357 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 261,356B, BPFP=1.4753 -⌛️ [2/4] FRONTEND: Frontend time: 0.414s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.output: torch.Size([1, 346, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.698s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.output: torch.Size([1, 346, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02507911 6.82317586 - layer.0.v_cache 0.00000027 0.00023337 - layer.1.k_cache 0.00290669 0.62240909 - layer.1.v_cache 0.00000082 0.00084860 - layer.2.k_cache 0.00115107 0.38453052 - layer.2.v_cache 0.00000117 0.00120339 - layer.3.k_cache 0.00130922 0.42849731 - layer.3.v_cache 0.00000214 0.00196337 - layer.4.k_cache 0.00347357 0.78175760 - layer.4.v_cache 0.00000325 0.00331074 - layer.4.output 0.00015455 0.06001676 - ------------------------------------------------------------------------------------- - TOTAL 0.00246754 0.66342835 - (elements=4,960,256) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4960256 -Total Bytes 977296 -BPFP 1.5762 bits/point -EBPFP 3.1524 equivalent bits/point -MSE 0.663428 ----------------------- -------------------------------------------------------- -Time: 1.130s Load: 0.018s, Pack+Encode: 0.414s, Decode+Unpack: 0.698s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 346, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6634 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample132-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample132-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample135-layer4-item1.zst (15/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample135-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 343, 128) -Output shape: (1, 343, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.output: torch.Size([1, 343, 4096]) -> torch.Size([1, 1, 343, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,104B, BPFP=0.6857 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,648B, BPFP=1.8369 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,332B, BPFP=1.3058 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,240B, BPFP=1.9871 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,812B, BPFP=1.4762 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,608B, BPFP=2.0182 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,904B, BPFP=1.5239 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,228B, BPFP=1.9868 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,252B, BPFP=1.3268 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,256B, BPFP=2.0330 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,516B, BPFP=1.4151 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.706s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02309080 6.62674600 - layer.0.v_cache 0.00000026 0.00021858 - layer.1.k_cache 0.00292563 0.62038126 - layer.1.v_cache 0.00000076 0.00083350 - layer.2.k_cache 0.00117950 0.39068639 - layer.2.v_cache 0.00000116 0.00115735 - layer.3.k_cache 0.00131686 0.43348645 - layer.3.v_cache 0.00000208 0.00183588 - layer.4.k_cache 0.00357410 0.76602048 - layer.4.v_cache 0.00000334 0.00322959 - layer.4.output 0.00016206 0.05255153 - ------------------------------------------------------------------------------------- - TOTAL 0.00233876 0.64677154 - (elements=4,917,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4917248 -Total Bytes 958900 -BPFP 1.5601 bits/point -EBPFP 3.1201 equivalent bits/point -MSE 0.646772 ----------------------- -------------------------------------------------------- -Time: 1.152s Load: 0.017s, Pack+Encode: 0.429s, Decode+Unpack: 0.706s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6468 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample135-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample135-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample14-layer4-item1.zst (16/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample14-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 359, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 359, 128) -Output shape: (1, 359, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.output: torch.Size([1, 359, 4096]) -> torch.Size([1, 1, 359, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,300B, BPFP=0.7029 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 84,904B, BPFP=1.8477 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,184B, BPFP=1.3097 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 92,296B, BPFP=2.0085 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,280B, BPFP=1.4859 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 93,588B, BPFP=2.0366 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 69,796B, BPFP=1.5189 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 92,024B, BPFP=2.0026 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,192B, BPFP=1.3317 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 93,840B, BPFP=2.0421 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 261,176B, BPFP=1.4209 -⌛️ [2/4] FRONTEND: Frontend time: 0.415s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.output: torch.Size([1, 359, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.701s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.output: torch.Size([1, 359, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02394788 6.89157232 - layer.0.v_cache 0.00000027 0.00022644 - layer.1.k_cache 0.00287336 0.63258307 - layer.1.v_cache 0.00000081 0.00090442 - layer.2.k_cache 0.00118844 0.39758416 - layer.2.v_cache 0.00000120 0.00125829 - layer.3.k_cache 0.00133454 0.43151753 - layer.3.v_cache 0.00000220 0.00197765 - layer.4.k_cache 0.00356152 0.76316702 - layer.4.v_cache 0.00000363 0.00344110 - layer.4.output 0.00015158 0.05255322 - ------------------------------------------------------------------------------------- - TOTAL 0.00239430 0.66674606 - (elements=5,146,624) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5146624 -Total Bytes 1009580 -BPFP 1.5693 bits/point -EBPFP 3.1386 equivalent bits/point -MSE 0.666746 ----------------------- -------------------------------------------------------- -Time: 1.134s Load: 0.018s, Pack+Encode: 0.415s, Decode+Unpack: 0.701s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 359, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6667 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample14-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample14-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample15-layer4-item1.zst (17/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample15-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 359, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 359, 128) -Output shape: (1, 359, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) -> torch.Size([1, 1, 359, 1024]) - layer.4.output: torch.Size([1, 359, 4096]) -> torch.Size([1, 1, 359, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,656B, BPFP=0.7107 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 85,152B, BPFP=1.8531 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 59,860B, BPFP=1.3027 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 91,884B, BPFP=1.9996 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,140B, BPFP=1.4829 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 93,388B, BPFP=2.0323 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,000B, BPFP=1.5233 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 91,896B, BPFP=1.9998 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,116B, BPFP=1.3300 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 93,924B, BPFP=2.0440 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 265,456B, BPFP=1.4442 -⌛️ [2/4] FRONTEND: Frontend time: 0.442s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.output: torch.Size([1, 359, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.709s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 359, 128]) - layer.0.v_cache: torch.Size([1, 8, 359, 128]) - layer.1.k_cache: torch.Size([1, 8, 359, 128]) - layer.1.v_cache: torch.Size([1, 8, 359, 128]) - layer.2.k_cache: torch.Size([1, 8, 359, 128]) - layer.2.v_cache: torch.Size([1, 8, 359, 128]) - layer.3.k_cache: torch.Size([1, 8, 359, 128]) - layer.3.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.k_cache: torch.Size([1, 8, 359, 128]) - layer.4.v_cache: torch.Size([1, 8, 359, 128]) - layer.4.output: torch.Size([1, 359, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02338462 6.78433406 - layer.0.v_cache 0.00000027 0.00023441 - layer.1.k_cache 0.00293092 0.62588824 - layer.1.v_cache 0.00000080 0.00087471 - layer.2.k_cache 0.00116855 0.39079535 - layer.2.v_cache 0.00000117 0.00124900 - layer.3.k_cache 0.00131991 0.43961948 - layer.3.v_cache 0.00000231 0.00207881 - layer.4.k_cache 0.00358756 0.76850640 - layer.4.v_cache 0.00000356 0.00350944 - layer.4.output 0.00016308 0.05406222 - ------------------------------------------------------------------------------------- - TOTAL 0.00236086 0.65952420 - (elements=5,146,624) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5146624 -Total Bytes 1013472 -BPFP 1.5754 bits/point -EBPFP 3.1507 equivalent bits/point -MSE 0.659524 ----------------------- -------------------------------------------------------- -Time: 1.168s Load: 0.017s, Pack+Encode: 0.442s, Decode+Unpack: 0.709s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 359, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 359, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6595 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample15-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample15-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample16-layer4-item1.zst (18/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample16-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 347, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 347, 128) -Output shape: (1, 347, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.0.v_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.1.k_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.1.v_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.2.k_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.2.v_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.3.k_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.3.v_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.4.k_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.4.v_cache: torch.Size([1, 8, 347, 128]) -> torch.Size([1, 1, 347, 1024]) - layer.4.output: torch.Size([1, 347, 4096]) -> torch.Size([1, 1, 347, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,772B, BPFP=0.7153 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,220B, BPFP=1.8511 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,868B, BPFP=1.3029 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,372B, BPFP=1.9896 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,664B, BPFP=1.4784 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,732B, BPFP=2.0203 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,400B, BPFP=1.5175 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,292B, BPFP=1.9878 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,168B, BPFP=1.3321 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,620B, BPFP=2.0403 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,164B, BPFP=1.4250 -⌛️ [2/4] FRONTEND: Frontend time: 0.446s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 347, 128]) - layer.0.v_cache: torch.Size([1, 8, 347, 128]) - layer.1.k_cache: torch.Size([1, 8, 347, 128]) - layer.1.v_cache: torch.Size([1, 8, 347, 128]) - layer.2.k_cache: torch.Size([1, 8, 347, 128]) - layer.2.v_cache: torch.Size([1, 8, 347, 128]) - layer.3.k_cache: torch.Size([1, 8, 347, 128]) - layer.3.v_cache: torch.Size([1, 8, 347, 128]) - layer.4.k_cache: torch.Size([1, 8, 347, 128]) - layer.4.v_cache: torch.Size([1, 8, 347, 128]) - layer.4.output: torch.Size([1, 347, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.722s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 347, 128]) - layer.0.v_cache: torch.Size([1, 8, 347, 128]) - layer.1.k_cache: torch.Size([1, 8, 347, 128]) - layer.1.v_cache: torch.Size([1, 8, 347, 128]) - layer.2.k_cache: torch.Size([1, 8, 347, 128]) - layer.2.v_cache: torch.Size([1, 8, 347, 128]) - layer.3.k_cache: torch.Size([1, 8, 347, 128]) - layer.3.v_cache: torch.Size([1, 8, 347, 128]) - layer.4.k_cache: torch.Size([1, 8, 347, 128]) - layer.4.v_cache: torch.Size([1, 8, 347, 128]) - layer.4.output: torch.Size([1, 347, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02341881 6.59067467 - layer.0.v_cache 0.00000026 0.00023707 - layer.1.k_cache 0.00293388 0.62742061 - layer.1.v_cache 0.00000079 0.00087277 - layer.2.k_cache 0.00115427 0.39021860 - layer.2.v_cache 0.00000115 0.00120606 - layer.3.k_cache 0.00129772 0.43329205 - layer.3.v_cache 0.00000217 0.00197854 - layer.4.k_cache 0.00353122 0.74824273 - layer.4.v_cache 0.00000331 0.00341989 - layer.4.output 0.00014441 0.05428051 - ------------------------------------------------------------------------------------- - TOTAL 0.00235152 0.64390607 - (elements=4,974,592) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4974592 -Total Bytes 974272 -BPFP 1.5668 bits/point -EBPFP 3.1336 equivalent bits/point -MSE 0.643906 ----------------------- -------------------------------------------------------- -Time: 1.185s Load: 0.017s, Pack+Encode: 0.446s, Decode+Unpack: 0.722s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 347, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 347, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6439 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample16-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample16-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample165-layer4-item1.zst (19/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample165-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 337, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 337, 128) -Output shape: (1, 337, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.output: torch.Size([1, 337, 4096]) -> torch.Size([1, 1, 337, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,320B, BPFP=0.6797 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,268B, BPFP=1.8376 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,900B, BPFP=1.2959 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,312B, BPFP=1.9546 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,120B, BPFP=1.4633 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,348B, BPFP=2.0018 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,660B, BPFP=1.5222 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,332B, BPFP=1.9782 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,964B, BPFP=1.3206 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,692B, BPFP=2.0329 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 249,512B, BPFP=1.4461 -⌛️ [2/4] FRONTEND: Frontend time: 0.425s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.output: torch.Size([1, 337, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.714s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.output: torch.Size([1, 337, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02403486 7.34045374 - layer.0.v_cache 0.00000026 0.00022815 - layer.1.k_cache 0.00293885 0.65345171 - layer.1.v_cache 0.00000076 0.00077835 - layer.2.k_cache 0.00113696 0.39151522 - layer.2.v_cache 0.00000113 0.00109858 - layer.3.k_cache 0.00132482 0.44868985 - layer.3.v_cache 0.00000208 0.00184952 - layer.4.k_cache 0.00376518 0.82006066 - layer.4.v_cache 0.00000304 0.00314856 - layer.4.output 0.00014776 0.05837871 - ------------------------------------------------------------------------------------- - TOTAL 0.00241421 0.70677066 - (elements=4,831,232) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4831232 -Total Bytes 943428 -BPFP 1.5622 bits/point -EBPFP 3.1244 equivalent bits/point -MSE 0.706771 ----------------------- -------------------------------------------------------- -Time: 1.156s Load: 0.017s, Pack+Encode: 0.425s, Decode+Unpack: 0.714s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 337, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7068 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample165-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample165-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample17-layer4-item1.zst (20/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample17-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 351, 128) -Output shape: (1, 351, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.output: torch.Size([1, 351, 4096]) -> torch.Size([1, 1, 351, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,792B, BPFP=0.6854 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,128B, BPFP=1.8280 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,132B, BPFP=1.2939 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,060B, BPFP=1.9600 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,700B, BPFP=1.4623 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,844B, BPFP=1.9997 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,804B, BPFP=1.5092 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,184B, BPFP=1.9628 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,276B, BPFP=1.3194 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,356B, BPFP=2.0111 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,716B, BPFP=1.4118 -⌛️ [2/4] FRONTEND: Frontend time: 0.419s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.713s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02526469 7.52103922 - layer.0.v_cache 0.00000026 0.00023638 - layer.1.k_cache 0.00294286 0.63676787 - layer.1.v_cache 0.00000076 0.00083047 - layer.2.k_cache 0.00116394 0.38495260 - layer.2.v_cache 0.00000112 0.00116731 - layer.3.k_cache 0.00133268 0.42878145 - layer.3.v_cache 0.00000207 0.00187406 - layer.4.k_cache 0.00360177 0.77105209 - layer.4.v_cache 0.00000315 0.00325575 - layer.4.output 0.00019540 0.05504005 - ------------------------------------------------------------------------------------- - TOTAL 0.00250678 0.71215124 - (elements=5,031,936) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5031936 -Total Bytes 973992 -BPFP 1.5485 bits/point -EBPFP 3.0970 equivalent bits/point -MSE 0.712151 ----------------------- -------------------------------------------------------- -Time: 1.149s Load: 0.017s, Pack+Encode: 0.419s, Decode+Unpack: 0.713s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7122 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample17-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample17-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample18-layer4-item1.zst (21/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample18-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 346, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 346, 128) -Output shape: (1, 346, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) -> torch.Size([1, 1, 346, 1024]) - layer.4.output: torch.Size([1, 346, 4096]) -> torch.Size([1, 1, 346, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,124B, BPFP=0.7028 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,044B, BPFP=1.8299 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,292B, BPFP=1.2936 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,844B, BPFP=1.9835 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,084B, BPFP=1.4696 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,584B, BPFP=2.0228 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,820B, BPFP=1.5088 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,360B, BPFP=1.9951 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,344B, BPFP=1.3174 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,348B, BPFP=2.0400 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 255,816B, BPFP=1.4440 -⌛️ [2/4] FRONTEND: Frontend time: 0.427s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.output: torch.Size([1, 346, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.723s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 346, 128]) - layer.0.v_cache: torch.Size([1, 8, 346, 128]) - layer.1.k_cache: torch.Size([1, 8, 346, 128]) - layer.1.v_cache: torch.Size([1, 8, 346, 128]) - layer.2.k_cache: torch.Size([1, 8, 346, 128]) - layer.2.v_cache: torch.Size([1, 8, 346, 128]) - layer.3.k_cache: torch.Size([1, 8, 346, 128]) - layer.3.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.k_cache: torch.Size([1, 8, 346, 128]) - layer.4.v_cache: torch.Size([1, 8, 346, 128]) - layer.4.output: torch.Size([1, 346, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02293222 6.72774722 - layer.0.v_cache 0.00000026 0.00022297 - layer.1.k_cache 0.00295457 0.62617347 - layer.1.v_cache 0.00000081 0.00087284 - layer.2.k_cache 0.00115565 0.38477687 - layer.2.v_cache 0.00000119 0.00125728 - layer.3.k_cache 0.00131728 0.43270949 - layer.3.v_cache 0.00000219 0.00198730 - layer.4.k_cache 0.00358690 0.78235154 - layer.4.v_cache 0.00000349 0.00338250 - layer.4.output 0.00015659 0.05150029 - ------------------------------------------------------------------------------------- - TOTAL 0.00232721 0.65482019 - (elements=4,960,256) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4960256 -Total Bytes 971660 -BPFP 1.5671 bits/point -EBPFP 3.1342 equivalent bits/point -MSE 0.654820 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.018s, Pack+Encode: 0.427s, Decode+Unpack: 0.723s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 346, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 346, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6548 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample18-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample18-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample19-layer4-item1.zst (22/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample19-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 340, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.021s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 340, 128) -Output shape: (1, 340, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.output: torch.Size([1, 340, 4096]) -> torch.Size([1, 1, 340, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,040B, BPFP=0.7132 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,872B, BPFP=1.8583 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,584B, BPFP=1.3002 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,272B, BPFP=2.0053 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,452B, BPFP=1.4810 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,864B, BPFP=2.0419 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,000B, BPFP=1.5165 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,300B, BPFP=2.0060 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,848B, BPFP=1.3292 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,364B, BPFP=2.0534 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 251,488B, BPFP=1.4447 -⌛️ [2/4] FRONTEND: Frontend time: 0.430s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.output: torch.Size([1, 340, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.710s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.output: torch.Size([1, 340, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02361420 7.02007123 - layer.0.v_cache 0.00000026 0.00023989 - layer.1.k_cache 0.00281368 0.62815247 - layer.1.v_cache 0.00000080 0.00087570 - layer.2.k_cache 0.00120688 0.39662628 - layer.2.v_cache 0.00000119 0.00127081 - layer.3.k_cache 0.00132000 0.43584285 - layer.3.v_cache 0.00000216 0.00201623 - layer.4.k_cache 0.00354241 0.76309698 - layer.4.v_cache 0.00000378 0.00350733 - layer.4.output 0.00015030 0.05533980 - ------------------------------------------------------------------------------------- - TOTAL 0.00236476 0.67664707 - (elements=4,874,240) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4874240 -Total Bytes 961084 -BPFP 1.5774 bits/point -EBPFP 3.1548 equivalent bits/point -MSE 0.676647 ----------------------- -------------------------------------------------------- -Time: 1.161s Load: 0.021s, Pack+Encode: 0.430s, Decode+Unpack: 0.710s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 340, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6766 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample19-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample19-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample2-layer4-item1.zst (23/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample2-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 344, 128) -Output shape: (1, 344, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.output: torch.Size([1, 344, 4096]) -> torch.Size([1, 1, 344, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,644B, BPFP=0.7187 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,340B, BPFP=1.8473 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,012B, BPFP=1.2948 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,348B, BPFP=1.9837 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,588B, BPFP=1.4668 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,168B, BPFP=2.0251 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,748B, BPFP=1.5159 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,076B, BPFP=2.0003 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,436B, BPFP=1.3271 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,256B, BPFP=2.0498 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 256,512B, BPFP=1.4564 -⌛️ [2/4] FRONTEND: Frontend time: 0.419s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.722s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02321494 6.74111300 - layer.0.v_cache 0.00000027 0.00021844 - layer.1.k_cache 0.00289516 0.60388317 - layer.1.v_cache 0.00000077 0.00081686 - layer.2.k_cache 0.00116570 0.38495978 - layer.2.v_cache 0.00000112 0.00117018 - layer.3.k_cache 0.00132982 0.42387372 - layer.3.v_cache 0.00000218 0.00190155 - layer.4.k_cache 0.00351457 0.75795835 - layer.4.v_cache 0.00000367 0.00347765 - layer.4.output 0.00015182 0.05642640 - ------------------------------------------------------------------------------------- - TOTAL 0.00233825 0.65321988 - (elements=4,931,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4931584 -Total Bytes 971128 -BPFP 1.5754 bits/point -EBPFP 3.1507 equivalent bits/point -MSE 0.653220 ----------------------- -------------------------------------------------------- -Time: 1.158s Load: 0.017s, Pack+Encode: 0.419s, Decode+Unpack: 0.722s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6532 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample2-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample2-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample20-layer4-item1.zst (24/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample20-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 337, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 337, 128) -Output shape: (1, 337, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) -> torch.Size([1, 1, 337, 1024]) - layer.4.output: torch.Size([1, 337, 4096]) -> torch.Size([1, 1, 337, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,252B, BPFP=0.7013 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,224B, BPFP=1.8598 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,852B, BPFP=1.2948 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,548B, BPFP=2.0064 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,424B, BPFP=1.4703 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,788B, BPFP=2.0351 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,528B, BPFP=1.5191 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,200B, BPFP=1.9983 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,184B, BPFP=1.3257 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,412B, BPFP=2.0496 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 246,584B, BPFP=1.4291 -⌛️ [2/4] FRONTEND: Frontend time: 0.440s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.output: torch.Size([1, 337, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.704s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 337, 128]) - layer.0.v_cache: torch.Size([1, 8, 337, 128]) - layer.1.k_cache: torch.Size([1, 8, 337, 128]) - layer.1.v_cache: torch.Size([1, 8, 337, 128]) - layer.2.k_cache: torch.Size([1, 8, 337, 128]) - layer.2.v_cache: torch.Size([1, 8, 337, 128]) - layer.3.k_cache: torch.Size([1, 8, 337, 128]) - layer.3.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.k_cache: torch.Size([1, 8, 337, 128]) - layer.4.v_cache: torch.Size([1, 8, 337, 128]) - layer.4.output: torch.Size([1, 337, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02435875 7.24301845 - layer.0.v_cache 0.00000026 0.00023465 - layer.1.k_cache 0.00286757 0.63340334 - layer.1.v_cache 0.00000078 0.00086645 - layer.2.k_cache 0.00117745 0.39646568 - layer.2.v_cache 0.00000112 0.00119833 - layer.3.k_cache 0.00130673 0.43821042 - layer.3.v_cache 0.00000209 0.00192784 - layer.4.k_cache 0.00356837 0.76038820 - layer.4.v_cache 0.00000322 0.00336864 - layer.4.output 0.00014638 0.05249345 - ------------------------------------------------------------------------------------- - TOTAL 0.00241942 0.69207541 - (elements=4,831,232) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4831232 -Total Bytes 947996 -BPFP 1.5698 bits/point -EBPFP 3.1396 equivalent bits/point -MSE 0.692075 ----------------------- -------------------------------------------------------- -Time: 1.162s Load: 0.018s, Pack+Encode: 0.440s, Decode+Unpack: 0.704s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 337, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 337, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6921 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample20-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample20-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample21-layer4-item1.zst (25/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample21-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 395, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.020s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 395, 128) -Output shape: (1, 395, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.0.v_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.1.k_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.1.v_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.2.k_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.2.v_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.3.k_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.3.v_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.4.k_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.4.v_cache: torch.Size([1, 8, 395, 128]) -> torch.Size([1, 1, 395, 1024]) - layer.4.output: torch.Size([1, 395, 4096]) -> torch.Size([1, 1, 395, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 35,016B, BPFP=0.6926 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 92,964B, BPFP=1.8387 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 65,244B, BPFP=1.2904 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 99,516B, BPFP=1.9683 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 73,108B, BPFP=1.4460 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 100,976B, BPFP=1.9972 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 76,372B, BPFP=1.5105 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 99,856B, BPFP=1.9750 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 66,472B, BPFP=1.3147 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 102,172B, BPFP=2.0208 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 291,896B, BPFP=1.4433 -⌛️ [2/4] FRONTEND: Frontend time: 0.589s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 395, 128]) - layer.0.v_cache: torch.Size([1, 8, 395, 128]) - layer.1.k_cache: torch.Size([1, 8, 395, 128]) - layer.1.v_cache: torch.Size([1, 8, 395, 128]) - layer.2.k_cache: torch.Size([1, 8, 395, 128]) - layer.2.v_cache: torch.Size([1, 8, 395, 128]) - layer.3.k_cache: torch.Size([1, 8, 395, 128]) - layer.3.v_cache: torch.Size([1, 8, 395, 128]) - layer.4.k_cache: torch.Size([1, 8, 395, 128]) - layer.4.v_cache: torch.Size([1, 8, 395, 128]) - layer.4.output: torch.Size([1, 395, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.827s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 395, 128]) - layer.0.v_cache: torch.Size([1, 8, 395, 128]) - layer.1.k_cache: torch.Size([1, 8, 395, 128]) - layer.1.v_cache: torch.Size([1, 8, 395, 128]) - layer.2.k_cache: torch.Size([1, 8, 395, 128]) - layer.2.v_cache: torch.Size([1, 8, 395, 128]) - layer.3.k_cache: torch.Size([1, 8, 395, 128]) - layer.3.v_cache: torch.Size([1, 8, 395, 128]) - layer.4.k_cache: torch.Size([1, 8, 395, 128]) - layer.4.v_cache: torch.Size([1, 8, 395, 128]) - layer.4.output: torch.Size([1, 395, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02333523 6.90850722 - layer.0.v_cache 0.00000027 0.00024301 - layer.1.k_cache 0.00291865 0.63159102 - layer.1.v_cache 0.00000080 0.00083778 - layer.2.k_cache 0.00114101 0.38734444 - layer.2.v_cache 0.00000113 0.00115455 - layer.3.k_cache 0.00133612 0.44026497 - layer.3.v_cache 0.00000221 0.00186999 - layer.4.k_cache 0.00365959 0.77669531 - layer.4.v_cache 0.00000309 0.00322382 - layer.4.output 0.00017803 0.06334982 - ------------------------------------------------------------------------------------- - TOTAL 0.00236502 0.67179510 - (elements=5,662,720) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5662720 -Total Bytes 1103592 -BPFP 1.5591 bits/point -EBPFP 3.1182 equivalent bits/point -MSE 0.671795 ----------------------- -------------------------------------------------------- -Time: 1.436s Load: 0.020s, Pack+Encode: 0.589s, Decode+Unpack: 0.827s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 395, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 395, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6718 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample21-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample21-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample22-layer4-item1.zst (26/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample22-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 348, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 348, 128) -Output shape: (1, 348, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.output: torch.Size([1, 348, 4096]) -> torch.Size([1, 1, 348, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,044B, BPFP=0.6969 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,176B, BPFP=1.8448 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,188B, BPFP=1.3063 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,340B, BPFP=1.9832 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,908B, BPFP=1.4796 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,924B, BPFP=2.0188 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,700B, BPFP=1.5198 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,100B, BPFP=1.9778 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,240B, BPFP=1.3299 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,552B, BPFP=2.0329 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 250,836B, BPFP=1.4078 -⌛️ [2/4] FRONTEND: Frontend time: 0.421s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.output: torch.Size([1, 348, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.726s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.output: torch.Size([1, 348, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02579716 6.87705064 - layer.0.v_cache 0.00000026 0.00023949 - layer.1.k_cache 0.00289751 0.63953479 - layer.1.v_cache 0.00000077 0.00087043 - layer.2.k_cache 0.00114660 0.38866828 - layer.2.v_cache 0.00000117 0.00123583 - layer.3.k_cache 0.00133042 0.43215329 - layer.3.v_cache 0.00000210 0.00193977 - layer.4.k_cache 0.00358099 0.76802335 - layer.4.v_cache 0.00000335 0.00342761 - layer.4.output 0.00013932 0.05627100 - ------------------------------------------------------------------------------------- - TOTAL 0.00252269 0.66701625 - (elements=4,988,928) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4988928 -Total Bytes 972008 -BPFP 1.5587 bits/point -EBPFP 3.1173 equivalent bits/point -MSE 0.667016 ----------------------- -------------------------------------------------------- -Time: 1.163s Load: 0.017s, Pack+Encode: 0.421s, Decode+Unpack: 0.726s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 348, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6670 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample22-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample22-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample23-layer4-item1.zst (27/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample23-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 316, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.020s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 316, 128) -Output shape: (1, 316, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.output: torch.Size([1, 316, 4096]) -> torch.Size([1, 1, 316, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,876B, BPFP=0.7139 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,116B, BPFP=1.8077 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,028B, BPFP=1.2863 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,904B, BPFP=1.9260 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,740B, BPFP=1.4522 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,172B, BPFP=1.9574 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,408B, BPFP=1.4935 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,184B, BPFP=1.9330 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,284B, BPFP=1.3173 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,748B, BPFP=1.9716 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 230,824B, BPFP=1.4267 -⌛️ [2/4] FRONTEND: Frontend time: 0.378s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.output: torch.Size([1, 316, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.609s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.output: torch.Size([1, 316, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02512236 6.38462926 - layer.0.v_cache 0.00000027 0.00022622 - layer.1.k_cache 0.00293329 0.62238404 - layer.1.v_cache 0.00000080 0.00083216 - layer.2.k_cache 0.00117467 0.39041567 - layer.2.v_cache 0.00000111 0.00118019 - layer.3.k_cache 0.00131319 0.42628397 - layer.3.v_cache 0.00000216 0.00194036 - layer.4.k_cache 0.00359311 0.74509217 - layer.4.v_cache 0.00000344 0.00336357 - layer.4.output 0.00015753 0.05681891 - ------------------------------------------------------------------------------------- - TOTAL 0.00248390 0.62883023 - (elements=4,530,176) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4530176 -Total Bytes 872284 -BPFP 1.5404 bits/point -EBPFP 3.0808 equivalent bits/point -MSE 0.628830 ----------------------- -------------------------------------------------------- -Time: 1.007s Load: 0.020s, Pack+Encode: 0.378s, Decode+Unpack: 0.609s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 316, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6288 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample23-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample23-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample24-layer4-item1.zst (28/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample24-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 348, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 348, 128) -Output shape: (1, 348, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) -> torch.Size([1, 1, 348, 1024]) - layer.4.output: torch.Size([1, 348, 4096]) -> torch.Size([1, 1, 348, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,184B, BPFP=0.7001 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,884B, BPFP=1.8383 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,252B, BPFP=1.3077 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,664B, BPFP=1.9905 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,680B, BPFP=1.4745 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,344B, BPFP=2.0282 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,620B, BPFP=1.5180 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,708B, BPFP=1.9915 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,384B, BPFP=1.3332 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,888B, BPFP=2.0404 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 255,196B, BPFP=1.4323 -⌛️ [2/4] FRONTEND: Frontend time: 0.423s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.output: torch.Size([1, 348, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.723s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 348, 128]) - layer.0.v_cache: torch.Size([1, 8, 348, 128]) - layer.1.k_cache: torch.Size([1, 8, 348, 128]) - layer.1.v_cache: torch.Size([1, 8, 348, 128]) - layer.2.k_cache: torch.Size([1, 8, 348, 128]) - layer.2.v_cache: torch.Size([1, 8, 348, 128]) - layer.3.k_cache: torch.Size([1, 8, 348, 128]) - layer.3.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.k_cache: torch.Size([1, 8, 348, 128]) - layer.4.v_cache: torch.Size([1, 8, 348, 128]) - layer.4.output: torch.Size([1, 348, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02368600 7.01315764 - layer.0.v_cache 0.00000027 0.00023827 - layer.1.k_cache 0.00288004 0.63986083 - layer.1.v_cache 0.00000079 0.00087428 - layer.2.k_cache 0.00117070 0.39441922 - layer.2.v_cache 0.00000116 0.00126215 - layer.3.k_cache 0.00131678 0.43579185 - layer.3.v_cache 0.00000214 0.00200036 - layer.4.k_cache 0.00357349 0.77695570 - layer.4.v_cache 0.00000347 0.00344216 - layer.4.output 0.00013697 0.05434397 - ------------------------------------------------------------------------------------- - TOTAL 0.00237019 0.67752703 - (elements=4,988,928) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4988928 -Total Bytes 977804 -BPFP 1.5680 bits/point -EBPFP 3.1359 equivalent bits/point -MSE 0.677527 ----------------------- -------------------------------------------------------- -Time: 1.164s Load: 0.018s, Pack+Encode: 0.423s, Decode+Unpack: 0.723s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 348, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 348, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6775 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample24-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample24-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample25-layer4-item1.zst (29/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample25-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 345, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.024s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 345, 128) -Output shape: (1, 345, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.output: torch.Size([1, 345, 4096]) -> torch.Size([1, 1, 345, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,888B, BPFP=0.6995 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,944B, BPFP=1.8330 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,440B, BPFP=1.3007 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,672B, BPFP=1.9853 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,348B, BPFP=1.4798 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,584B, BPFP=2.0286 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,148B, BPFP=1.5206 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,160B, BPFP=1.9964 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,504B, BPFP=1.3248 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,180B, BPFP=2.0421 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 254,124B, BPFP=1.4387 -⌛️ [2/4] FRONTEND: Frontend time: 0.431s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.output: torch.Size([1, 345, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.717s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.output: torch.Size([1, 345, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02357864 6.70243787 - layer.0.v_cache 0.00000026 0.00022398 - layer.1.k_cache 0.00288937 0.62087986 - layer.1.v_cache 0.00000077 0.00085652 - layer.2.k_cache 0.00116899 0.39102668 - layer.2.v_cache 0.00000115 0.00123945 - layer.3.k_cache 0.00131620 0.42561968 - layer.3.v_cache 0.00000216 0.00197576 - layer.4.k_cache 0.00353981 0.77962248 - layer.4.v_cache 0.00000338 0.00342520 - layer.4.output 0.00014567 0.05036048 - ------------------------------------------------------------------------------------- - TOTAL 0.00236310 0.65205353 - (elements=4,945,920) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4945920 -Total Bytes 969992 -BPFP 1.5690 bits/point -EBPFP 3.1379 equivalent bits/point -MSE 0.652054 ----------------------- -------------------------------------------------------- -Time: 1.172s Load: 0.024s, Pack+Encode: 0.431s, Decode+Unpack: 0.717s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 345, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6521 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample25-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample25-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample26-layer4-item1.zst (30/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample26-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 330, 128) -Output shape: (1, 330, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.output: torch.Size([1, 330, 4096]) -> torch.Size([1, 1, 330, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,088B, BPFP=0.7360 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,360B, BPFP=1.8788 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,240B, BPFP=1.3078 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,196B, BPFP=2.0170 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,776B, BPFP=1.4862 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,980B, BPFP=2.0592 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,772B, BPFP=1.5334 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,664B, BPFP=2.0280 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,324B, BPFP=1.3334 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,616B, BPFP=2.0742 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,868B, BPFP=1.4729 -⌛️ [2/4] FRONTEND: Frontend time: 0.424s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.697s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02434325 6.94338305 - layer.0.v_cache 0.00000027 0.00023607 - layer.1.k_cache 0.00287412 0.63004164 - layer.1.v_cache 0.00000081 0.00085431 - layer.2.k_cache 0.00119273 0.39847606 - layer.2.v_cache 0.00000117 0.00124486 - layer.3.k_cache 0.00130297 0.42998986 - layer.3.v_cache 0.00000221 0.00201352 - layer.4.k_cache 0.00352071 0.76982477 - layer.4.v_cache 0.00000330 0.00349979 - layer.4.output 0.00014370 0.05568764 - ------------------------------------------------------------------------------------- - TOTAL 0.00241545 0.67159389 - (elements=4,730,880) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4730880 -Total Bytes 943884 -BPFP 1.5961 bits/point -EBPFP 3.1922 equivalent bits/point -MSE 0.671594 ----------------------- -------------------------------------------------------- -Time: 1.140s Load: 0.019s, Pack+Encode: 0.424s, Decode+Unpack: 0.697s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6716 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample26-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample26-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample27-layer4-item1.zst (31/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample27-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 334, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.023s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 334, 128) -Output shape: (1, 334, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.output: torch.Size([1, 334, 4096]) -> torch.Size([1, 1, 334, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,840B, BPFP=0.7214 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,692B, BPFP=1.8641 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,832B, BPFP=1.3060 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,568B, BPFP=2.0015 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,924B, BPFP=1.4718 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,564B, BPFP=2.0482 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,224B, BPFP=1.5256 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,908B, BPFP=2.0094 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,988B, BPFP=1.3330 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,908B, BPFP=2.0562 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 250,820B, BPFP=1.4667 -⌛️ [2/4] FRONTEND: Frontend time: 0.446s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.output: torch.Size([1, 334, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.720s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.output: torch.Size([1, 334, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02363260 7.02942699 - layer.0.v_cache 0.00000026 0.00023392 - layer.1.k_cache 0.00290221 0.65071179 - layer.1.v_cache 0.00000079 0.00085490 - layer.2.k_cache 0.00116170 0.39092332 - layer.2.v_cache 0.00000115 0.00124181 - layer.3.k_cache 0.00132769 0.44613076 - layer.3.v_cache 0.00000211 0.00192916 - layer.4.k_cache 0.00362300 0.78761630 - layer.4.v_cache 0.00000352 0.00343989 - layer.4.output 0.00014243 0.06033560 - ------------------------------------------------------------------------------------- - TOTAL 0.00237320 0.68241795 - (elements=4,788,224) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4788224 -Total Bytes 949268 -BPFP 1.5860 bits/point -EBPFP 3.1720 equivalent bits/point -MSE 0.682418 ----------------------- -------------------------------------------------------- -Time: 1.188s Load: 0.023s, Pack+Encode: 0.446s, Decode+Unpack: 0.720s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 334, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6824 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample27-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample27-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample28-layer4-item1.zst (32/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample28-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 330, 128) -Output shape: (1, 330, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.output: torch.Size([1, 330, 4096]) -> torch.Size([1, 1, 330, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,620B, BPFP=0.7249 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,880B, BPFP=1.8674 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,148B, BPFP=1.3056 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,896B, BPFP=2.0098 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,636B, BPFP=1.4829 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,672B, BPFP=2.0519 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,712B, BPFP=1.5320 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,376B, BPFP=2.0212 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,228B, BPFP=1.3312 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,332B, BPFP=2.0675 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 249,928B, BPFP=1.4792 -⌛️ [2/4] FRONTEND: Frontend time: 0.439s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.717s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02410887 6.82490974 - layer.0.v_cache 0.00000026 0.00023695 - layer.1.k_cache 0.00288475 0.62639082 - layer.1.v_cache 0.00000079 0.00085210 - layer.2.k_cache 0.00119186 0.39430408 - layer.2.v_cache 0.00000115 0.00122073 - layer.3.k_cache 0.00132245 0.43363490 - layer.3.v_cache 0.00000226 0.00201481 - layer.4.k_cache 0.00352035 0.78277153 - layer.4.v_cache 0.00000339 0.00340010 - layer.4.output 0.00014978 0.06422565 - ------------------------------------------------------------------------------------- - TOTAL 0.00240252 0.66618846 - (elements=4,730,880) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4730880 -Total Bytes 942428 -BPFP 1.5937 bits/point -EBPFP 3.1873 equivalent bits/point -MSE 0.666188 ----------------------- -------------------------------------------------------- -Time: 1.174s Load: 0.018s, Pack+Encode: 0.439s, Decode+Unpack: 0.717s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6662 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample28-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample28-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample29-layer4-item1.zst (33/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample29-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 329, 128) -Output shape: (1, 329, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.output: torch.Size([1, 329, 4096]) -> torch.Size([1, 1, 329, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,828B, BPFP=0.7320 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,156B, BPFP=1.8797 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,724B, BPFP=1.2995 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,600B, BPFP=2.0089 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,300B, BPFP=1.4794 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,816B, BPFP=2.0378 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,508B, BPFP=1.5318 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,828B, BPFP=2.0143 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,296B, BPFP=1.3368 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,180B, BPFP=2.0702 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,724B, BPFP=1.4766 -⌛️ [2/4] FRONTEND: Frontend time: 0.420s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.712s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02426959 7.19167636 - layer.0.v_cache 0.00000027 0.00023247 - layer.1.k_cache 0.00291730 0.64580394 - layer.1.v_cache 0.00000079 0.00082815 - layer.2.k_cache 0.00116725 0.39037891 - layer.2.v_cache 0.00000113 0.00116515 - layer.3.k_cache 0.00130327 0.43296281 - layer.3.v_cache 0.00000218 0.00192975 - layer.4.k_cache 0.00359671 0.75236878 - layer.4.v_cache 0.00000327 0.00344581 - layer.4.output 0.00014627 0.05746004 - ------------------------------------------------------------------------------------- - TOTAL 0.00241763 0.68933088 - (elements=4,716,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4716544 -Total Bytes 938960 -BPFP 1.5926 bits/point -EBPFP 3.1852 equivalent bits/point -MSE 0.689331 ----------------------- -------------------------------------------------------- -Time: 1.149s Load: 0.017s, Pack+Encode: 0.420s, Decode+Unpack: 0.712s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6893 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample29-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample29-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample3-layer4-item1.zst (34/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample3-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 365, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 365, 128) -Output shape: (1, 365, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.0.v_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.1.k_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.1.v_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.2.k_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.2.v_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.3.k_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.3.v_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.4.k_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.4.v_cache: torch.Size([1, 8, 365, 128]) -> torch.Size([1, 1, 365, 1024]) - layer.4.output: torch.Size([1, 365, 4096]) -> torch.Size([1, 1, 365, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,920B, BPFP=0.7046 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 86,324B, BPFP=1.8477 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,560B, BPFP=1.2962 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 92,760B, BPFP=1.9854 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,876B, BPFP=1.4742 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 94,476B, BPFP=2.0222 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,388B, BPFP=1.5066 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 93,008B, BPFP=1.9908 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,840B, BPFP=1.3236 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 94,772B, BPFP=2.0285 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 271,332B, BPFP=1.4519 -⌛️ [2/4] FRONTEND: Frontend time: 0.433s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 365, 128]) - layer.0.v_cache: torch.Size([1, 8, 365, 128]) - layer.1.k_cache: torch.Size([1, 8, 365, 128]) - layer.1.v_cache: torch.Size([1, 8, 365, 128]) - layer.2.k_cache: torch.Size([1, 8, 365, 128]) - layer.2.v_cache: torch.Size([1, 8, 365, 128]) - layer.3.k_cache: torch.Size([1, 8, 365, 128]) - layer.3.v_cache: torch.Size([1, 8, 365, 128]) - layer.4.k_cache: torch.Size([1, 8, 365, 128]) - layer.4.v_cache: torch.Size([1, 8, 365, 128]) - layer.4.output: torch.Size([1, 365, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.728s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 365, 128]) - layer.0.v_cache: torch.Size([1, 8, 365, 128]) - layer.1.k_cache: torch.Size([1, 8, 365, 128]) - layer.1.v_cache: torch.Size([1, 8, 365, 128]) - layer.2.k_cache: torch.Size([1, 8, 365, 128]) - layer.2.v_cache: torch.Size([1, 8, 365, 128]) - layer.3.k_cache: torch.Size([1, 8, 365, 128]) - layer.3.v_cache: torch.Size([1, 8, 365, 128]) - layer.4.k_cache: torch.Size([1, 8, 365, 128]) - layer.4.v_cache: torch.Size([1, 8, 365, 128]) - layer.4.output: torch.Size([1, 365, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02487199 6.81106392 - layer.0.v_cache 0.00000027 0.00023634 - layer.1.k_cache 0.00290279 0.62310879 - layer.1.v_cache 0.00000079 0.00086774 - layer.2.k_cache 0.00116885 0.39525895 - layer.2.v_cache 0.00000114 0.00125330 - layer.3.k_cache 0.00131452 0.43574177 - layer.3.v_cache 0.00000216 0.00200601 - layer.4.k_cache 0.00360833 0.75329841 - layer.4.v_cache 0.00000353 0.00338455 - layer.4.output 0.00014951 0.05472227 - ------------------------------------------------------------------------------------- - TOTAL 0.00246232 0.66036492 - (elements=5,232,640) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5232640 -Total Bytes 1027256 -BPFP 1.5705 bits/point -EBPFP 3.1411 equivalent bits/point -MSE 0.660365 ----------------------- -------------------------------------------------------- -Time: 1.181s Load: 0.019s, Pack+Encode: 0.433s, Decode+Unpack: 0.728s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 365, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 365, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6604 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample3-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample3-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample30-layer4-item1.zst (35/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample30-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 335, 128) -Output shape: (1, 335, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.output: torch.Size([1, 335, 4096]) -> torch.Size([1, 1, 335, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,996B, BPFP=0.6995 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,032B, BPFP=1.8664 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,084B, BPFP=1.3079 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,596B, BPFP=2.0195 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,408B, BPFP=1.4787 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,856B, BPFP=2.0489 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,376B, BPFP=1.5246 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,276B, BPFP=2.0120 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,036B, BPFP=1.3301 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,224B, BPFP=2.0575 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 249,500B, BPFP=1.4546 -⌛️ [2/4] FRONTEND: Frontend time: 0.417s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02462482 6.55343182 - layer.0.v_cache 0.00000027 0.00023360 - layer.1.k_cache 0.00287179 0.62471345 - layer.1.v_cache 0.00000079 0.00087582 - layer.2.k_cache 0.00118919 0.39210141 - layer.2.v_cache 0.00000114 0.00123114 - layer.3.k_cache 0.00130718 0.42894852 - layer.3.v_cache 0.00000215 0.00193599 - layer.4.k_cache 0.00355403 0.76615810 - layer.4.v_cache 0.00000335 0.00334350 - layer.4.output 0.00013864 0.05330220 - ------------------------------------------------------------------------------------- - TOTAL 0.00243638 0.64187015 - (elements=4,802,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4802560 -Total Bytes 950384 -BPFP 1.5831 bits/point -EBPFP 3.1663 equivalent bits/point -MSE 0.641870 ----------------------- -------------------------------------------------------- -Time: 1.141s Load: 0.017s, Pack+Encode: 0.417s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6419 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample30-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample30-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample31-layer4-item1.zst (36/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample31-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.023s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 335, 128) -Output shape: (1, 335, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.output: torch.Size([1, 335, 4096]) -> torch.Size([1, 1, 335, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,072B, BPFP=0.7013 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,652B, BPFP=1.8576 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,956B, BPFP=1.3049 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,876B, BPFP=2.0027 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,404B, BPFP=1.4786 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,748B, BPFP=2.0464 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,212B, BPFP=1.5208 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,164B, BPFP=2.0094 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,680B, BPFP=1.3218 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,088B, BPFP=2.0543 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,796B, BPFP=1.4505 -⌛️ [2/4] FRONTEND: Frontend time: 0.437s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.719s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02394044 6.84814745 - layer.0.v_cache 0.00000026 0.00022834 - layer.1.k_cache 0.00289851 0.64398635 - layer.1.v_cache 0.00000079 0.00084676 - layer.2.k_cache 0.00115963 0.39481051 - layer.2.v_cache 0.00000116 0.00125474 - layer.3.k_cache 0.00131431 0.43454758 - layer.3.v_cache 0.00000219 0.00201237 - layer.4.k_cache 0.00369786 0.78271238 - layer.4.v_cache 0.00000335 0.00338067 - layer.4.output 0.00014933 0.05981411 - ------------------------------------------------------------------------------------- - TOTAL 0.00240113 0.66794168 - (elements=4,802,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4802560 -Total Bytes 947648 -BPFP 1.5786 bits/point -EBPFP 3.1571 equivalent bits/point -MSE 0.667942 ----------------------- -------------------------------------------------------- -Time: 1.178s Load: 0.023s, Pack+Encode: 0.437s, Decode+Unpack: 0.719s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6679 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample31-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample31-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample32-layer4-item1.zst (37/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample32-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 339, 128) -Output shape: (1, 339, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.output: torch.Size([1, 339, 4096]) -> torch.Size([1, 1, 339, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,000B, BPFP=0.7144 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,724B, BPFP=1.8603 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,824B, BPFP=1.3096 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,784B, BPFP=2.0000 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,320B, BPFP=1.4823 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,940B, BPFP=2.0497 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,148B, BPFP=1.5244 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,400B, BPFP=2.0142 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,652B, BPFP=1.3286 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,220B, BPFP=2.0561 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,484B, BPFP=1.4604 -⌛️ [2/4] FRONTEND: Frontend time: 0.417s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.703s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02382689 7.01686515 - layer.0.v_cache 0.00000027 0.00023872 - layer.1.k_cache 0.00292607 0.63409716 - layer.1.v_cache 0.00000081 0.00084888 - layer.2.k_cache 0.00125336 0.39382066 - layer.2.v_cache 0.00000116 0.00123513 - layer.3.k_cache 0.00131849 0.42460736 - layer.3.v_cache 0.00000220 0.00200323 - layer.4.k_cache 0.00356446 0.75139233 - layer.4.v_cache 0.00000343 0.00339302 - layer.4.output 0.00016600 0.05808160 - ------------------------------------------------------------------------------------- - TOTAL 0.00239722 0.67577343 - (elements=4,859,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4859904 -Total Bytes 962496 -BPFP 1.5844 bits/point -EBPFP 3.1688 equivalent bits/point -MSE 0.675773 ----------------------- -------------------------------------------------------- -Time: 1.138s Load: 0.018s, Pack+Encode: 0.417s, Decode+Unpack: 0.703s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6758 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample32-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample32-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample33-layer4-item1.zst (38/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample33-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 362, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 362, 128) -Output shape: (1, 362, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) -> torch.Size([1, 1, 362, 1024]) - layer.4.output: torch.Size([1, 362, 4096]) -> torch.Size([1, 1, 362, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,852B, BPFP=0.6874 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 85,836B, BPFP=1.8525 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,216B, BPFP=1.2996 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 92,504B, BPFP=1.9964 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,504B, BPFP=1.4784 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 93,828B, BPFP=2.0249 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,180B, BPFP=1.5146 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 92,196B, BPFP=1.9897 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,608B, BPFP=1.3296 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 94,336B, BPFP=2.0359 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 263,040B, BPFP=1.4192 -⌛️ [2/4] FRONTEND: Frontend time: 0.431s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.output: torch.Size([1, 362, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.717s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 362, 128]) - layer.0.v_cache: torch.Size([1, 8, 362, 128]) - layer.1.k_cache: torch.Size([1, 8, 362, 128]) - layer.1.v_cache: torch.Size([1, 8, 362, 128]) - layer.2.k_cache: torch.Size([1, 8, 362, 128]) - layer.2.v_cache: torch.Size([1, 8, 362, 128]) - layer.3.k_cache: torch.Size([1, 8, 362, 128]) - layer.3.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.k_cache: torch.Size([1, 8, 362, 128]) - layer.4.v_cache: torch.Size([1, 8, 362, 128]) - layer.4.output: torch.Size([1, 362, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02338883 7.19057064 - layer.0.v_cache 0.00000028 0.00024033 - layer.1.k_cache 0.00287711 0.62967526 - layer.1.v_cache 0.00000077 0.00086183 - layer.2.k_cache 0.00119138 0.39511366 - layer.2.v_cache 0.00000112 0.00120697 - layer.3.k_cache 0.00131510 0.42704680 - layer.3.v_cache 0.00000211 0.00194057 - layer.4.k_cache 0.00352384 0.75728999 - layer.4.v_cache 0.00000335 0.00335745 - layer.4.output 0.00013081 0.05151856 - ------------------------------------------------------------------------------------- - TOTAL 0.00234480 0.68666984 - (elements=5,189,632) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5189632 -Total Bytes 1014100 -BPFP 1.5633 bits/point -EBPFP 3.1265 equivalent bits/point -MSE 0.686670 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.019s, Pack+Encode: 0.431s, Decode+Unpack: 0.717s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 362, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 362, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6867 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample33-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample33-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample34-layer4-item1.zst (39/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample34-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 318, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.016s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 318, 128) -Output shape: (1, 318, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) -> torch.Size([1, 1, 318, 1024]) - layer.4.output: torch.Size([1, 318, 4096]) -> torch.Size([1, 1, 318, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,748B, BPFP=0.7308 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,248B, BPFP=1.7995 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,512B, BPFP=1.2901 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,284B, BPFP=1.9233 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,932B, BPFP=1.4478 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,576B, BPFP=1.9550 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,568B, BPFP=1.4880 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,640B, BPFP=1.9320 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 54,032B, BPFP=1.3274 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 80,144B, BPFP=1.9689 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 237,180B, BPFP=1.4567 -⌛️ [2/4] FRONTEND: Frontend time: 0.369s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.output: torch.Size([1, 318, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.611s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 318, 128]) - layer.0.v_cache: torch.Size([1, 8, 318, 128]) - layer.1.k_cache: torch.Size([1, 8, 318, 128]) - layer.1.v_cache: torch.Size([1, 8, 318, 128]) - layer.2.k_cache: torch.Size([1, 8, 318, 128]) - layer.2.v_cache: torch.Size([1, 8, 318, 128]) - layer.3.k_cache: torch.Size([1, 8, 318, 128]) - layer.3.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.k_cache: torch.Size([1, 8, 318, 128]) - layer.4.v_cache: torch.Size([1, 8, 318, 128]) - layer.4.output: torch.Size([1, 318, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02477463 6.74325926 - layer.0.v_cache 0.00000027 0.00023249 - layer.1.k_cache 0.00289381 0.62112758 - layer.1.v_cache 0.00000083 0.00087066 - layer.2.k_cache 0.00119129 0.39882646 - layer.2.v_cache 0.00000117 0.00127266 - layer.3.k_cache 0.00131631 0.44164295 - layer.3.v_cache 0.00000257 0.00204799 - layer.4.k_cache 0.00356231 0.77016008 - layer.4.v_cache 0.00000346 0.00348245 - layer.4.output 0.00015646 0.06145027 - ------------------------------------------------------------------------------------- - TOTAL 0.00245518 0.65919455 - (elements=4,558,848) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4558848 -Total Bytes 882864 -BPFP 1.5493 bits/point -EBPFP 3.0986 equivalent bits/point -MSE 0.659195 ----------------------- -------------------------------------------------------- -Time: 0.996s Load: 0.016s, Pack+Encode: 0.369s, Decode+Unpack: 0.611s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 318, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 318, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6592 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample34-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample34-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample35-layer4-item1.zst (40/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample35-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 342, 128) -Output shape: (1, 342, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.output: torch.Size([1, 342, 4096]) -> torch.Size([1, 1, 342, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,640B, BPFP=0.6999 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,820B, BPFP=1.8462 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,072B, BPFP=1.3037 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,516B, BPFP=1.9992 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,540B, BPFP=1.4743 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,104B, BPFP=2.0355 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,576B, BPFP=1.5208 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,460B, BPFP=1.9979 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,016B, BPFP=1.3253 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,584B, BPFP=2.0464 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,052B, BPFP=1.4452 -⌛️ [2/4] FRONTEND: Frontend time: 0.426s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.718s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02487079 6.72516904 - layer.0.v_cache 0.00000026 0.00022871 - layer.1.k_cache 0.00288341 0.62208994 - layer.1.v_cache 0.00000082 0.00085143 - layer.2.k_cache 0.00117177 0.39011981 - layer.2.v_cache 0.00000117 0.00122576 - layer.3.k_cache 0.00130255 0.43308535 - layer.3.v_cache 0.00000215 0.00193742 - layer.4.k_cache 0.00351771 0.74983197 - layer.4.v_cache 0.00000337 0.00337611 - layer.4.output 0.00013439 0.05144979 - ------------------------------------------------------------------------------------- - TOTAL 0.00244940 0.65240819 - (elements=4,902,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4902912 -Total Bytes 964380 -BPFP 1.5736 bits/point -EBPFP 3.1471 equivalent bits/point -MSE 0.652408 ----------------------- -------------------------------------------------------- -Time: 1.162s Load: 0.017s, Pack+Encode: 0.426s, Decode+Unpack: 0.718s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6524 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample35-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample35-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample36-layer4-item1.zst (41/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample36-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 343, 128) -Output shape: (1, 343, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.output: torch.Size([1, 343, 4096]) -> torch.Size([1, 1, 343, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,320B, BPFP=0.7134 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,780B, BPFP=1.8627 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,276B, BPFP=1.3046 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,972B, BPFP=2.0037 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,052B, BPFP=1.4817 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,544B, BPFP=2.0395 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,528B, BPFP=1.5153 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,028B, BPFP=2.0050 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,452B, BPFP=1.3314 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,012B, BPFP=2.0502 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 255,492B, BPFP=1.4548 -⌛️ [2/4] FRONTEND: Frontend time: 0.424s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.721s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02787456 6.74027993 - layer.0.v_cache 0.00000027 0.00023028 - layer.1.k_cache 0.00288980 0.60568108 - layer.1.v_cache 0.00000078 0.00086350 - layer.2.k_cache 0.00118800 0.39104895 - layer.2.v_cache 0.00000116 0.00122095 - layer.3.k_cache 0.00131625 0.42750523 - layer.3.v_cache 0.00000215 0.00199296 - layer.4.k_cache 0.00354145 0.75247175 - layer.4.v_cache 0.00000355 0.00342205 - layer.4.output 0.00016614 0.05586891 - ------------------------------------------------------------------------------------- - TOTAL 0.00267732 0.65344231 - (elements=4,917,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4917248 -Total Bytes 971456 -BPFP 1.5805 bits/point -EBPFP 3.1610 equivalent bits/point -MSE 0.653442 ----------------------- -------------------------------------------------------- -Time: 1.162s Load: 0.017s, Pack+Encode: 0.424s, Decode+Unpack: 0.721s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6534 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample36-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample36-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample37-layer4-item1.zst (42/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample37-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 333, 128) -Output shape: (1, 333, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.output: torch.Size([1, 333, 4096]) -> torch.Size([1, 1, 333, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,212B, BPFP=0.7088 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,660B, BPFP=1.8689 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,752B, BPFP=1.3080 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,656B, BPFP=2.0096 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,924B, BPFP=1.4763 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,276B, BPFP=2.0476 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,748B, BPFP=1.5191 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,700B, BPFP=2.0106 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,692B, BPFP=1.3300 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,772B, BPFP=2.0592 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 244,648B, BPFP=1.4349 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.716s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02325115 6.98345636 - layer.0.v_cache 0.00000026 0.00023223 - layer.1.k_cache 0.00292822 0.65769037 - layer.1.v_cache 0.00000083 0.00085958 - layer.2.k_cache 0.00115284 0.39165339 - layer.2.v_cache 0.00000118 0.00123771 - layer.3.k_cache 0.00132958 0.43545853 - layer.3.v_cache 0.00000212 0.00194726 - layer.4.k_cache 0.00351398 0.76063373 - layer.4.v_cache 0.00000364 0.00341296 - layer.4.output 0.00014544 0.05731415 - ------------------------------------------------------------------------------------- - TOTAL 0.00234040 0.67613134 - (elements=4,773,888) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4773888 -Total Bytes 941040 -BPFP 1.5770 bits/point -EBPFP 3.1540 equivalent bits/point -MSE 0.676131 ----------------------- -------------------------------------------------------- -Time: 1.163s Load: 0.018s, Pack+Encode: 0.429s, Decode+Unpack: 0.716s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6761 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample37-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample37-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample38-layer4-item1.zst (43/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample38-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 356, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 356, 128) -Output shape: (1, 356, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.output: torch.Size([1, 356, 4096]) -> torch.Size([1, 1, 356, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,376B, BPFP=0.6886 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 83,540B, BPFP=1.8333 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 59,248B, BPFP=1.3002 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 90,508B, BPFP=1.9862 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 67,320B, BPFP=1.4774 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 92,144B, BPFP=2.0221 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 69,068B, BPFP=1.5157 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 90,816B, BPFP=1.9930 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 60,244B, BPFP=1.3221 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 92,908B, BPFP=2.0389 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 263,368B, BPFP=1.4449 -⌛️ [2/4] FRONTEND: Frontend time: 0.425s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.output: torch.Size([1, 356, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.713s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.output: torch.Size([1, 356, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02485108 7.33001229 - layer.0.v_cache 0.00000027 0.00023440 - layer.1.k_cache 0.00289550 0.64375451 - layer.1.v_cache 0.00000079 0.00086585 - layer.2.k_cache 0.00117641 0.40098499 - layer.2.v_cache 0.00000113 0.00123425 - layer.3.k_cache 0.00131994 0.43964780 - layer.3.v_cache 0.00000217 0.00200475 - layer.4.k_cache 0.00362269 0.77794167 - layer.4.v_cache 0.00000343 0.00343990 - layer.4.output 0.00016041 0.05701777 - ------------------------------------------------------------------------------------- - TOTAL 0.00246536 0.70201368 - (elements=5,103,616) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5103616 -Total Bytes 1000540 -BPFP 1.5684 bits/point -EBPFP 3.1367 equivalent bits/point -MSE 0.702014 ----------------------- -------------------------------------------------------- -Time: 1.157s Load: 0.018s, Pack+Encode: 0.425s, Decode+Unpack: 0.713s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 356, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7020 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample38-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample38-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample39-layer4-item1.zst (44/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample39-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 331, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 331, 128) -Output shape: (1, 331, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.0.v_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.1.k_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.1.v_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.2.k_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.2.v_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.3.k_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.3.v_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.4.k_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.4.v_cache: torch.Size([1, 8, 331, 128]) -> torch.Size([1, 1, 331, 1024]) - layer.4.output: torch.Size([1, 331, 4096]) -> torch.Size([1, 1, 331, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,676B, BPFP=0.7240 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,148B, BPFP=1.8681 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,516B, BPFP=1.3103 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,300B, BPFP=2.0133 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,740B, BPFP=1.4808 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,864B, BPFP=2.0502 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,908B, BPFP=1.5320 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,108B, BPFP=2.0088 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,564B, BPFP=1.3351 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,288B, BPFP=2.0602 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 246,812B, BPFP=1.4564 -⌛️ [2/4] FRONTEND: Frontend time: 0.425s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 331, 128]) - layer.0.v_cache: torch.Size([1, 8, 331, 128]) - layer.1.k_cache: torch.Size([1, 8, 331, 128]) - layer.1.v_cache: torch.Size([1, 8, 331, 128]) - layer.2.k_cache: torch.Size([1, 8, 331, 128]) - layer.2.v_cache: torch.Size([1, 8, 331, 128]) - layer.3.k_cache: torch.Size([1, 8, 331, 128]) - layer.3.v_cache: torch.Size([1, 8, 331, 128]) - layer.4.k_cache: torch.Size([1, 8, 331, 128]) - layer.4.v_cache: torch.Size([1, 8, 331, 128]) - layer.4.output: torch.Size([1, 331, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.704s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 331, 128]) - layer.0.v_cache: torch.Size([1, 8, 331, 128]) - layer.1.k_cache: torch.Size([1, 8, 331, 128]) - layer.1.v_cache: torch.Size([1, 8, 331, 128]) - layer.2.k_cache: torch.Size([1, 8, 331, 128]) - layer.2.v_cache: torch.Size([1, 8, 331, 128]) - layer.3.k_cache: torch.Size([1, 8, 331, 128]) - layer.3.v_cache: torch.Size([1, 8, 331, 128]) - layer.4.k_cache: torch.Size([1, 8, 331, 128]) - layer.4.v_cache: torch.Size([1, 8, 331, 128]) - layer.4.output: torch.Size([1, 331, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02326727 6.56739904 - layer.0.v_cache 0.00000026 0.00022724 - layer.1.k_cache 0.00291335 0.63650600 - layer.1.v_cache 0.00000080 0.00085870 - layer.2.k_cache 0.00116981 0.39991719 - layer.2.v_cache 0.00000115 0.00123093 - layer.3.k_cache 0.00132745 0.43479200 - layer.3.v_cache 0.00000216 0.00193572 - layer.4.k_cache 0.00359183 0.77829643 - layer.4.v_cache 0.00000354 0.00335257 - layer.4.output 0.00014916 0.05930287 - ------------------------------------------------------------------------------------- - TOTAL 0.00234816 0.64726624 - (elements=4,745,216) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4745216 -Total Bytes 940924 -BPFP 1.5863 bits/point -EBPFP 3.1726 equivalent bits/point -MSE 0.647266 ----------------------- -------------------------------------------------------- -Time: 1.146s Load: 0.017s, Pack+Encode: 0.425s, Decode+Unpack: 0.704s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 331, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 331, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6473 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample39-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample39-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample4-layer4-item1.zst (45/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample4-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 345, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 345, 128) -Output shape: (1, 345, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) -> torch.Size([1, 1, 345, 1024]) - layer.4.output: torch.Size([1, 345, 4096]) -> torch.Size([1, 1, 345, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,444B, BPFP=0.7120 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,004B, BPFP=1.8343 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,640B, BPFP=1.2826 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,376B, BPFP=1.9560 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,324B, BPFP=1.4566 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,684B, BPFP=2.0082 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,676B, BPFP=1.5099 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,664B, BPFP=1.9851 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,836B, BPFP=1.3097 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,596B, BPFP=2.0289 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 259,528B, BPFP=1.4692 -⌛️ [2/4] FRONTEND: Frontend time: 0.434s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.output: torch.Size([1, 345, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.720s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 345, 128]) - layer.0.v_cache: torch.Size([1, 8, 345, 128]) - layer.1.k_cache: torch.Size([1, 8, 345, 128]) - layer.1.v_cache: torch.Size([1, 8, 345, 128]) - layer.2.k_cache: torch.Size([1, 8, 345, 128]) - layer.2.v_cache: torch.Size([1, 8, 345, 128]) - layer.3.k_cache: torch.Size([1, 8, 345, 128]) - layer.3.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.k_cache: torch.Size([1, 8, 345, 128]) - layer.4.v_cache: torch.Size([1, 8, 345, 128]) - layer.4.output: torch.Size([1, 345, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02346668 6.90059655 - layer.0.v_cache 0.00000026 0.00022769 - layer.1.k_cache 0.00293293 0.62633110 - layer.1.v_cache 0.00000076 0.00079515 - layer.2.k_cache 0.00115776 0.37658665 - layer.2.v_cache 0.00000114 0.00119713 - layer.3.k_cache 0.00132774 0.43124633 - layer.3.v_cache 0.00000222 0.00195872 - layer.4.k_cache 0.00355356 0.75111084 - layer.4.v_cache 0.00000324 0.00328208 - layer.4.output 0.00016540 0.05917756 - ------------------------------------------------------------------------------------- - TOTAL 0.00236485 0.66643161 - (elements=4,945,920) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4945920 -Total Bytes 969772 -BPFP 1.5686 bits/point -EBPFP 3.1372 equivalent bits/point -MSE 0.666432 ----------------------- -------------------------------------------------------- -Time: 1.172s Load: 0.018s, Pack+Encode: 0.434s, Decode+Unpack: 0.720s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 345, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 345, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6664 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample4-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample4-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample40-layer4-item1.zst (46/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample40-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 344, 128) -Output shape: (1, 344, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.output: torch.Size([1, 344, 4096]) -> torch.Size([1, 1, 344, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,380B, BPFP=0.7127 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,232B, BPFP=1.8448 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,308B, BPFP=1.3015 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,408B, BPFP=1.9851 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,692B, BPFP=1.4692 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,268B, BPFP=2.0273 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,996B, BPFP=1.5215 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,976B, BPFP=1.9980 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,492B, BPFP=1.3284 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,004B, BPFP=2.0441 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 255,600B, BPFP=1.4512 -⌛️ [2/4] FRONTEND: Frontend time: 0.430s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.719s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02380104 7.07050554 - layer.0.v_cache 0.00000026 0.00022833 - layer.1.k_cache 0.00287155 0.61011598 - layer.1.v_cache 0.00000078 0.00084229 - layer.2.k_cache 0.00117748 0.38938434 - layer.2.v_cache 0.00000114 0.00118465 - layer.3.k_cache 0.00132414 0.41940152 - layer.3.v_cache 0.00000215 0.00191083 - layer.4.k_cache 0.00352424 0.75277666 - layer.4.v_cache 0.00000349 0.00333977 - layer.4.output 0.00015726 0.05390159 - ------------------------------------------------------------------------------------- - TOTAL 0.00238109 0.67609259 - (elements=4,931,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4931584 -Total Bytes 970356 -BPFP 1.5741 bits/point -EBPFP 3.1482 equivalent bits/point -MSE 0.676093 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.019s, Pack+Encode: 0.430s, Decode+Unpack: 0.719s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6761 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample40-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample40-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample41-layer4-item1.zst (47/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample41-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 339, 128) -Output shape: (1, 339, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.output: torch.Size([1, 339, 4096]) -> torch.Size([1, 1, 339, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,728B, BPFP=0.7081 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,876B, BPFP=1.8638 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,508B, BPFP=1.3023 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,056B, BPFP=2.0063 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,104B, BPFP=1.4773 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,508B, BPFP=2.0397 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,944B, BPFP=1.5197 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,000B, BPFP=2.0050 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,724B, BPFP=1.3303 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,976B, BPFP=2.0505 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 252,116B, BPFP=1.4525 -⌛️ [2/4] FRONTEND: Frontend time: 0.421s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.725s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02519707 6.73609047 - layer.0.v_cache 0.00000026 0.00023367 - layer.1.k_cache 0.00285639 0.62673302 - layer.1.v_cache 0.00000078 0.00086648 - layer.2.k_cache 0.00115969 0.39374103 - layer.2.v_cache 0.00000117 0.00121093 - layer.3.k_cache 0.00131371 0.42690171 - layer.3.v_cache 0.00000219 0.00194029 - layer.4.k_cache 0.00358480 0.77101140 - layer.4.v_cache 0.00000371 0.00339136 - layer.4.output 0.00015316 0.05945301 - ------------------------------------------------------------------------------------- - TOTAL 0.00248089 0.65713803 - (elements=4,859,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4859904 -Total Bytes 959540 -BPFP 1.5795 bits/point -EBPFP 3.1590 equivalent bits/point -MSE 0.657138 ----------------------- -------------------------------------------------------- -Time: 1.164s Load: 0.018s, Pack+Encode: 0.421s, Decode+Unpack: 0.725s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6571 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample41-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample41-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample42-layer4-item1.zst (48/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample42-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 316, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 316, 128) -Output shape: (1, 316, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) -> torch.Size([1, 1, 316, 1024]) - layer.4.output: torch.Size([1, 316, 4096]) -> torch.Size([1, 1, 316, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,836B, BPFP=0.7129 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,180B, BPFP=1.8092 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,276B, BPFP=1.2924 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,048B, BPFP=1.9296 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,856B, BPFP=1.4551 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,540B, BPFP=1.9665 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,572B, BPFP=1.4975 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,548B, BPFP=1.9420 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,544B, BPFP=1.3238 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 80,304B, BPFP=1.9854 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 234,900B, BPFP=1.4519 -⌛️ [2/4] FRONTEND: Frontend time: 0.379s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.output: torch.Size([1, 316, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.601s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 316, 128]) - layer.0.v_cache: torch.Size([1, 8, 316, 128]) - layer.1.k_cache: torch.Size([1, 8, 316, 128]) - layer.1.v_cache: torch.Size([1, 8, 316, 128]) - layer.2.k_cache: torch.Size([1, 8, 316, 128]) - layer.2.v_cache: torch.Size([1, 8, 316, 128]) - layer.3.k_cache: torch.Size([1, 8, 316, 128]) - layer.3.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.k_cache: torch.Size([1, 8, 316, 128]) - layer.4.v_cache: torch.Size([1, 8, 316, 128]) - layer.4.output: torch.Size([1, 316, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02408213 6.65969269 - layer.0.v_cache 0.00000026 0.00022873 - layer.1.k_cache 0.00292725 0.61319187 - layer.1.v_cache 0.00000082 0.00084570 - layer.2.k_cache 0.00117417 0.38734574 - layer.2.v_cache 0.00000116 0.00123481 - layer.3.k_cache 0.00131215 0.43528274 - layer.3.v_cache 0.00000217 0.00199859 - layer.4.k_cache 0.00352248 0.74983070 - layer.4.v_cache 0.00000354 0.00347796 - layer.4.output 0.00015546 0.06249077 - ------------------------------------------------------------------------------------- - TOTAL 0.00240343 0.65022090 - (elements=4,530,176) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4530176 -Total Bytes 878604 -BPFP 1.5516 bits/point -EBPFP 3.1031 equivalent bits/point -MSE 0.650221 ----------------------- -------------------------------------------------------- -Time: 0.997s Load: 0.018s, Pack+Encode: 0.379s, Decode+Unpack: 0.601s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 316, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 316, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6502 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample42-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample42-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample43-layer4-item1.zst (49/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample43-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 343, 128) -Output shape: (1, 343, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) -> torch.Size([1, 1, 343, 1024]) - layer.4.output: torch.Size([1, 343, 4096]) -> torch.Size([1, 1, 343, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,760B, BPFP=0.7006 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,412B, BPFP=1.8543 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,348B, BPFP=1.3062 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,800B, BPFP=1.9998 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,984B, BPFP=1.4801 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,440B, BPFP=2.0372 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,776B, BPFP=1.5210 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,796B, BPFP=1.9997 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,320B, BPFP=1.3284 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,908B, BPFP=2.0478 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,208B, BPFP=1.4418 -⌛️ [2/4] FRONTEND: Frontend time: 0.415s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.718s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 343, 128]) - layer.0.v_cache: torch.Size([1, 8, 343, 128]) - layer.1.k_cache: torch.Size([1, 8, 343, 128]) - layer.1.v_cache: torch.Size([1, 8, 343, 128]) - layer.2.k_cache: torch.Size([1, 8, 343, 128]) - layer.2.v_cache: torch.Size([1, 8, 343, 128]) - layer.3.k_cache: torch.Size([1, 8, 343, 128]) - layer.3.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.k_cache: torch.Size([1, 8, 343, 128]) - layer.4.v_cache: torch.Size([1, 8, 343, 128]) - layer.4.output: torch.Size([1, 343, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02365684 6.70401473 - layer.0.v_cache 0.00000027 0.00023186 - layer.1.k_cache 0.00286131 0.62414898 - layer.1.v_cache 0.00000077 0.00086385 - layer.2.k_cache 0.00115688 0.39729082 - layer.2.v_cache 0.00000135 0.00123861 - layer.3.k_cache 0.00129955 0.43124648 - layer.3.v_cache 0.00000214 0.00195197 - layer.4.k_cache 0.00368945 0.76260002 - layer.4.v_cache 0.00000337 0.00340973 - layer.4.output 0.00015649 0.05287924 - ------------------------------------------------------------------------------------- - TOTAL 0.00237842 0.65275100 - (elements=4,917,248) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4917248 -Total Bytes 967752 -BPFP 1.5745 bits/point -EBPFP 3.1489 equivalent bits/point -MSE 0.652751 ----------------------- -------------------------------------------------------- -Time: 1.151s Load: 0.017s, Pack+Encode: 0.415s, Decode+Unpack: 0.718s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 343, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 343, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6528 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample43-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample43-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample44-layer4-item1.zst (50/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample44-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 326, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 326, 128) -Output shape: (1, 326, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.0.v_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.1.k_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.1.v_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.2.k_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.2.v_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.3.k_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.3.v_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.4.k_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.4.v_cache: torch.Size([1, 8, 326, 128]) -> torch.Size([1, 1, 326, 1024]) - layer.4.output: torch.Size([1, 326, 4096]) -> torch.Size([1, 1, 326, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,844B, BPFP=0.7152 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,060B, BPFP=1.8707 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 53,904B, BPFP=1.2918 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 83,376B, BPFP=1.9981 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,308B, BPFP=1.4692 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,060B, BPFP=2.0384 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 63,904B, BPFP=1.5314 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 83,872B, BPFP=2.0100 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,200B, BPFP=1.3229 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,032B, BPFP=2.0617 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 242,336B, BPFP=1.4519 -⌛️ [2/4] FRONTEND: Frontend time: 0.422s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 326, 128]) - layer.0.v_cache: torch.Size([1, 8, 326, 128]) - layer.1.k_cache: torch.Size([1, 8, 326, 128]) - layer.1.v_cache: torch.Size([1, 8, 326, 128]) - layer.2.k_cache: torch.Size([1, 8, 326, 128]) - layer.2.v_cache: torch.Size([1, 8, 326, 128]) - layer.3.k_cache: torch.Size([1, 8, 326, 128]) - layer.3.v_cache: torch.Size([1, 8, 326, 128]) - layer.4.k_cache: torch.Size([1, 8, 326, 128]) - layer.4.v_cache: torch.Size([1, 8, 326, 128]) - layer.4.output: torch.Size([1, 326, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.703s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 326, 128]) - layer.0.v_cache: torch.Size([1, 8, 326, 128]) - layer.1.k_cache: torch.Size([1, 8, 326, 128]) - layer.1.v_cache: torch.Size([1, 8, 326, 128]) - layer.2.k_cache: torch.Size([1, 8, 326, 128]) - layer.2.v_cache: torch.Size([1, 8, 326, 128]) - layer.3.k_cache: torch.Size([1, 8, 326, 128]) - layer.3.v_cache: torch.Size([1, 8, 326, 128]) - layer.4.k_cache: torch.Size([1, 8, 326, 128]) - layer.4.v_cache: torch.Size([1, 8, 326, 128]) - layer.4.output: torch.Size([1, 326, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02438200 6.93664775 - layer.0.v_cache 0.00000027 0.00023325 - layer.1.k_cache 0.00290854 0.65388063 - layer.1.v_cache 0.00000077 0.00082156 - layer.2.k_cache 0.00114310 0.38679018 - layer.2.v_cache 0.00000112 0.00118304 - layer.3.k_cache 0.00131875 0.43657811 - layer.3.v_cache 0.00000210 0.00188394 - layer.4.k_cache 0.00360283 0.77765997 - layer.4.v_cache 0.00000318 0.00326492 - layer.4.output 0.00014710 0.05450280 - ------------------------------------------------------------------------------------- - TOTAL 0.00242508 0.67263961 - (elements=4,673,536) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4673536 -Total Bytes 922896 -BPFP 1.5798 bits/point -EBPFP 3.1596 equivalent bits/point -MSE 0.672640 ----------------------- -------------------------------------------------------- -Time: 1.143s Load: 0.018s, Pack+Encode: 0.422s, Decode+Unpack: 0.703s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 326, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 326, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6726 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample44-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample44-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample45-layer4-item1.zst (51/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample45-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 335, 128) -Output shape: (1, 335, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.output: torch.Size([1, 335, 4096]) -> torch.Size([1, 1, 335, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,100B, BPFP=0.7253 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,144B, BPFP=1.8690 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,028B, BPFP=1.3066 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,212B, BPFP=2.0105 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,376B, BPFP=1.4780 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,644B, BPFP=2.0439 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,412B, BPFP=1.5255 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,248B, BPFP=2.0114 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,056B, BPFP=1.3306 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,216B, BPFP=2.0573 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 252,452B, BPFP=1.4719 -⌛️ [2/4] FRONTEND: Frontend time: 0.444s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.718s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02422888 7.12703329 - layer.0.v_cache 0.00000026 0.00023283 - layer.1.k_cache 0.00291837 0.65406007 - layer.1.v_cache 0.00000081 0.00088149 - layer.2.k_cache 0.00116761 0.39549866 - layer.2.v_cache 0.00000118 0.00125347 - layer.3.k_cache 0.00133299 0.43876775 - layer.3.v_cache 0.00000217 0.00200942 - layer.4.k_cache 0.00352932 0.78171642 - layer.4.v_cache 0.00000340 0.00340274 - layer.4.output 0.00015509 0.05659389 - ------------------------------------------------------------------------------------- - TOTAL 0.00241467 0.68794512 - (elements=4,802,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4802560 -Total Bytes 953888 -BPFP 1.5890 bits/point -EBPFP 3.1779 equivalent bits/point -MSE 0.687945 ----------------------- -------------------------------------------------------- -Time: 1.180s Load: 0.018s, Pack+Encode: 0.444s, Decode+Unpack: 0.718s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6879 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample45-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample45-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample46-layer4-item1.zst (52/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample46-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 324, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 324, 128) -Output shape: (1, 324, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.0.v_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.1.k_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.1.v_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.2.k_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.2.v_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.3.k_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.3.v_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.4.k_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.4.v_cache: torch.Size([1, 8, 324, 128]) -> torch.Size([1, 1, 324, 1024]) - layer.4.output: torch.Size([1, 324, 4096]) -> torch.Size([1, 1, 324, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,716B, BPFP=0.7165 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 77,204B, BPFP=1.8616 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,324B, BPFP=1.3099 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 83,376B, BPFP=2.0104 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,652B, BPFP=1.4866 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,176B, BPFP=2.0538 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 63,400B, BPFP=1.5287 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 83,752B, BPFP=2.0195 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,192B, BPFP=1.3308 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 85,768B, BPFP=2.0681 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 241,864B, BPFP=1.4580 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 324, 128]) - layer.0.v_cache: torch.Size([1, 8, 324, 128]) - layer.1.k_cache: torch.Size([1, 8, 324, 128]) - layer.1.v_cache: torch.Size([1, 8, 324, 128]) - layer.2.k_cache: torch.Size([1, 8, 324, 128]) - layer.2.v_cache: torch.Size([1, 8, 324, 128]) - layer.3.k_cache: torch.Size([1, 8, 324, 128]) - layer.3.v_cache: torch.Size([1, 8, 324, 128]) - layer.4.k_cache: torch.Size([1, 8, 324, 128]) - layer.4.v_cache: torch.Size([1, 8, 324, 128]) - layer.4.output: torch.Size([1, 324, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.695s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 324, 128]) - layer.0.v_cache: torch.Size([1, 8, 324, 128]) - layer.1.k_cache: torch.Size([1, 8, 324, 128]) - layer.1.v_cache: torch.Size([1, 8, 324, 128]) - layer.2.k_cache: torch.Size([1, 8, 324, 128]) - layer.2.v_cache: torch.Size([1, 8, 324, 128]) - layer.3.k_cache: torch.Size([1, 8, 324, 128]) - layer.3.v_cache: torch.Size([1, 8, 324, 128]) - layer.4.k_cache: torch.Size([1, 8, 324, 128]) - layer.4.v_cache: torch.Size([1, 8, 324, 128]) - layer.4.output: torch.Size([1, 324, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02313130 7.45366377 - layer.0.v_cache 0.00000026 0.00022884 - layer.1.k_cache 0.00295515 0.64737320 - layer.1.v_cache 0.00000079 0.00083587 - layer.2.k_cache 0.00117180 0.39385748 - layer.2.v_cache 0.00000115 0.00120995 - layer.3.k_cache 0.00132196 0.43720980 - layer.3.v_cache 0.00000215 0.00196221 - layer.4.k_cache 0.00355989 0.76578861 - layer.4.v_cache 0.00000355 0.00337007 - layer.4.output 0.00015612 0.05916133 - ------------------------------------------------------------------------------------- - TOTAL 0.00234089 0.71015322 - (elements=4,644,864) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4644864 -Total Bytes 921424 -BPFP 1.5870 bits/point -EBPFP 3.1740 equivalent bits/point -MSE 0.710153 ----------------------- -------------------------------------------------------- -Time: 1.128s Load: 0.017s, Pack+Encode: 0.416s, Decode+Unpack: 0.695s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 324, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 324, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7102 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample46-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample46-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample47-layer4-item1.zst (53/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample47-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 332, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 332, 128) -Output shape: (1, 332, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.0.v_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.1.k_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.1.v_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.2.k_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.2.v_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.3.k_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.3.v_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.4.k_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.4.v_cache: torch.Size([1, 8, 332, 128]) -> torch.Size([1, 1, 332, 1024]) - layer.4.output: torch.Size([1, 332, 4096]) -> torch.Size([1, 1, 332, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,760B, BPFP=0.7238 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,180B, BPFP=1.8632 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,460B, BPFP=1.3051 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,520B, BPFP=2.0124 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,928B, BPFP=1.4808 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,308B, BPFP=2.0545 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,016B, BPFP=1.5299 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,812B, BPFP=2.0193 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,488B, BPFP=1.3293 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,760B, BPFP=2.0651 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,952B, BPFP=1.4646 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 332, 128]) - layer.0.v_cache: torch.Size([1, 8, 332, 128]) - layer.1.k_cache: torch.Size([1, 8, 332, 128]) - layer.1.v_cache: torch.Size([1, 8, 332, 128]) - layer.2.k_cache: torch.Size([1, 8, 332, 128]) - layer.2.v_cache: torch.Size([1, 8, 332, 128]) - layer.3.k_cache: torch.Size([1, 8, 332, 128]) - layer.3.v_cache: torch.Size([1, 8, 332, 128]) - layer.4.k_cache: torch.Size([1, 8, 332, 128]) - layer.4.v_cache: torch.Size([1, 8, 332, 128]) - layer.4.output: torch.Size([1, 332, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.721s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 332, 128]) - layer.0.v_cache: torch.Size([1, 8, 332, 128]) - layer.1.k_cache: torch.Size([1, 8, 332, 128]) - layer.1.v_cache: torch.Size([1, 8, 332, 128]) - layer.2.k_cache: torch.Size([1, 8, 332, 128]) - layer.2.v_cache: torch.Size([1, 8, 332, 128]) - layer.3.k_cache: torch.Size([1, 8, 332, 128]) - layer.3.v_cache: torch.Size([1, 8, 332, 128]) - layer.4.k_cache: torch.Size([1, 8, 332, 128]) - layer.4.v_cache: torch.Size([1, 8, 332, 128]) - layer.4.output: torch.Size([1, 332, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02360621 7.11646979 - layer.0.v_cache 0.00000026 0.00022994 - layer.1.k_cache 0.00288168 0.64567952 - layer.1.v_cache 0.00000079 0.00087015 - layer.2.k_cache 0.00118772 0.40171874 - layer.2.v_cache 0.00000117 0.00129415 - layer.3.k_cache 0.00132341 0.43525912 - layer.3.v_cache 0.00000218 0.00202926 - layer.4.k_cache 0.00354949 0.76780862 - layer.4.v_cache 0.00000378 0.00350679 - layer.4.output 0.00018345 0.05787228 - ------------------------------------------------------------------------------------- - TOTAL 0.00237789 0.68616823 - (elements=4,759,552) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4759552 -Total Bytes 945184 -BPFP 1.5887 bits/point -EBPFP 3.1774 equivalent bits/point -MSE 0.686168 ----------------------- -------------------------------------------------------- -Time: 1.156s Load: 0.019s, Pack+Encode: 0.416s, Decode+Unpack: 0.721s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 332, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 332, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6862 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample47-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample47-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample48-layer4-item1.zst (54/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample48-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 319, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 319, 128) -Output shape: (1, 319, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.output: torch.Size([1, 319, 4096]) -> torch.Size([1, 1, 319, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,332B, BPFP=0.7184 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,084B, BPFP=1.7899 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 51,956B, BPFP=1.2724 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,188B, BPFP=1.9149 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,316B, BPFP=1.4282 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,364B, BPFP=1.9437 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 59,996B, BPFP=1.4693 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,960B, BPFP=1.9093 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,036B, BPFP=1.2989 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,676B, BPFP=1.9513 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 228,692B, BPFP=1.4002 -⌛️ [2/4] FRONTEND: Frontend time: 0.364s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.output: torch.Size([1, 319, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.592s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.output: torch.Size([1, 319, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02438342 6.44490611 - layer.0.v_cache 0.00000027 0.00023151 - layer.1.k_cache 0.00287626 0.59868911 - layer.1.v_cache 0.00000082 0.00086879 - layer.2.k_cache 0.00122126 0.38414296 - layer.2.v_cache 0.00000114 0.00122307 - layer.3.k_cache 0.00133159 0.43021735 - layer.3.v_cache 0.00000213 0.00192194 - layer.4.k_cache 0.00352266 0.76351226 - layer.4.v_cache 0.00000325 0.00334993 - layer.4.output 0.00013715 0.05004378 - ------------------------------------------------------------------------------------- - TOTAL 0.00242081 0.63065987 - (elements=4,573,184) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4573184 -Total Bytes 869600 -BPFP 1.5212 bits/point -EBPFP 3.0424 equivalent bits/point -MSE 0.630660 ----------------------- -------------------------------------------------------- -Time: 0.974s Load: 0.017s, Pack+Encode: 0.364s, Decode+Unpack: 0.592s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 319, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6307 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample48-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample48-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample49-layer4-item1.zst (55/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample49-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.022s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 335, 128) -Output shape: (1, 335, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) -> torch.Size([1, 1, 335, 1024]) - layer.4.output: torch.Size([1, 335, 4096]) -> torch.Size([1, 1, 335, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,436B, BPFP=0.7098 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,044B, BPFP=1.8667 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,732B, BPFP=1.2997 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,068B, BPFP=2.0072 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,344B, BPFP=1.4772 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,356B, BPFP=2.0372 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,156B, BPFP=1.5195 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,060B, BPFP=2.0070 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,164B, BPFP=1.3331 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,192B, BPFP=2.0567 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 246,460B, BPFP=1.4369 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.726s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 335, 128]) - layer.0.v_cache: torch.Size([1, 8, 335, 128]) - layer.1.k_cache: torch.Size([1, 8, 335, 128]) - layer.1.v_cache: torch.Size([1, 8, 335, 128]) - layer.2.k_cache: torch.Size([1, 8, 335, 128]) - layer.2.v_cache: torch.Size([1, 8, 335, 128]) - layer.3.k_cache: torch.Size([1, 8, 335, 128]) - layer.3.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.k_cache: torch.Size([1, 8, 335, 128]) - layer.4.v_cache: torch.Size([1, 8, 335, 128]) - layer.4.output: torch.Size([1, 335, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02322828 6.88038494 - layer.0.v_cache 0.00000026 0.00022762 - layer.1.k_cache 0.00283722 0.62894424 - layer.1.v_cache 0.00000078 0.00085730 - layer.2.k_cache 0.00117890 0.39089069 - layer.2.v_cache 0.00000113 0.00119702 - layer.3.k_cache 0.00131429 0.43112889 - layer.3.v_cache 0.00000212 0.00194435 - layer.4.k_cache 0.00357066 0.77268176 - layer.4.v_cache 0.00000341 0.00339159 - layer.4.output 0.00013949 0.05328286 - ------------------------------------------------------------------------------------- - TOTAL 0.00233536 0.66605570 - (elements=4,802,560) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4802560 -Total Bytes 946012 -BPFP 1.5758 bits/point -EBPFP 3.1517 equivalent bits/point -MSE 0.666056 ----------------------- -------------------------------------------------------- -Time: 1.178s Load: 0.022s, Pack+Encode: 0.429s, Decode+Unpack: 0.726s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 335, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 335, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6661 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample49-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample49-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample5-layer4-item1.zst (56/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample5-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 354, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 354, 128) -Output shape: (1, 354, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.0.v_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.1.k_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.1.v_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.2.k_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.2.v_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.3.k_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.3.v_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.4.k_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.4.v_cache: torch.Size([1, 8, 354, 128]) -> torch.Size([1, 1, 354, 1024]) - layer.4.output: torch.Size([1, 354, 4096]) -> torch.Size([1, 1, 354, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,608B, BPFP=0.7196 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 83,664B, BPFP=1.8464 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,912B, BPFP=1.2781 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 89,736B, BPFP=1.9804 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 66,024B, BPFP=1.4571 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 91,412B, BPFP=2.0174 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 68,072B, BPFP=1.5023 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 90,044B, BPFP=1.9872 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,852B, BPFP=1.3209 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 92,320B, BPFP=2.0374 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 262,192B, BPFP=1.4466 -⌛️ [2/4] FRONTEND: Frontend time: 0.426s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 354, 128]) - layer.0.v_cache: torch.Size([1, 8, 354, 128]) - layer.1.k_cache: torch.Size([1, 8, 354, 128]) - layer.1.v_cache: torch.Size([1, 8, 354, 128]) - layer.2.k_cache: torch.Size([1, 8, 354, 128]) - layer.2.v_cache: torch.Size([1, 8, 354, 128]) - layer.3.k_cache: torch.Size([1, 8, 354, 128]) - layer.3.v_cache: torch.Size([1, 8, 354, 128]) - layer.4.k_cache: torch.Size([1, 8, 354, 128]) - layer.4.v_cache: torch.Size([1, 8, 354, 128]) - layer.4.output: torch.Size([1, 354, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.725s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 354, 128]) - layer.0.v_cache: torch.Size([1, 8, 354, 128]) - layer.1.k_cache: torch.Size([1, 8, 354, 128]) - layer.1.v_cache: torch.Size([1, 8, 354, 128]) - layer.2.k_cache: torch.Size([1, 8, 354, 128]) - layer.2.v_cache: torch.Size([1, 8, 354, 128]) - layer.3.k_cache: torch.Size([1, 8, 354, 128]) - layer.3.v_cache: torch.Size([1, 8, 354, 128]) - layer.4.k_cache: torch.Size([1, 8, 354, 128]) - layer.4.v_cache: torch.Size([1, 8, 354, 128]) - layer.4.output: torch.Size([1, 354, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02475441 6.61033639 - layer.0.v_cache 0.00000027 0.00024160 - layer.1.k_cache 0.00288960 0.62555880 - layer.1.v_cache 0.00000085 0.00088569 - layer.2.k_cache 0.00116959 0.39910945 - layer.2.v_cache 0.00000115 0.00123459 - layer.3.k_cache 0.00132243 0.43816686 - layer.3.v_cache 0.00000218 0.00201230 - layer.4.k_cache 0.00355194 0.77375587 - layer.4.v_cache 0.00000329 0.00348910 - layer.4.output 0.00015235 0.06148909 - ------------------------------------------------------------------------------------- - TOTAL 0.00245037 0.65005336 - (elements=5,074,944) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5074944 -Total Bytes 993836 -BPFP 1.5667 bits/point -EBPFP 3.1333 equivalent bits/point -MSE 0.650053 ----------------------- -------------------------------------------------------- -Time: 1.170s Load: 0.019s, Pack+Encode: 0.426s, Decode+Unpack: 0.725s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 354, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 354, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6501 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample5-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample5-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample50-layer4-item1.zst (57/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample50-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 327, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.016s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 327, 128) -Output shape: (1, 327, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.output: torch.Size([1, 327, 4096]) -> torch.Size([1, 1, 327, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,548B, BPFP=0.7298 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,632B, BPFP=1.8786 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,780B, BPFP=1.3088 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,264B, BPFP=2.0132 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,040B, BPFP=1.4822 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,924B, BPFP=2.0528 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,272B, BPFP=1.5356 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,612B, BPFP=2.0215 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,636B, BPFP=1.3292 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,736B, BPFP=2.0722 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,588B, BPFP=1.4848 -⌛️ [2/4] FRONTEND: Frontend time: 0.410s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.output: torch.Size([1, 327, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.696s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.output: torch.Size([1, 327, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02504845 7.05537214 - layer.0.v_cache 0.00000027 0.00022689 - layer.1.k_cache 0.00290162 0.63840282 - layer.1.v_cache 0.00000083 0.00085252 - layer.2.k_cache 0.00115956 0.39162699 - layer.2.v_cache 0.00000120 0.00122747 - layer.3.k_cache 0.00131802 0.43236222 - layer.3.v_cache 0.00000223 0.00193648 - layer.4.k_cache 0.00351161 0.76886332 - layer.4.v_cache 0.00000355 0.00337364 - layer.4.output 0.00016630 0.05828290 - ------------------------------------------------------------------------------------- - TOTAL 0.00247233 0.68052686 - (elements=4,687,872) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4687872 -Total Bytes 936032 -BPFP 1.5974 bits/point -EBPFP 3.1947 equivalent bits/point -MSE 0.680527 ----------------------- -------------------------------------------------------- -Time: 1.123s Load: 0.016s, Pack+Encode: 0.410s, Decode+Unpack: 0.696s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 327, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6805 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample50-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample50-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample51-layer4-item1.zst (58/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample51-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 313, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.021s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 313, 128) -Output shape: (1, 313, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.0.v_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.1.k_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.1.v_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.2.k_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.2.v_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.3.k_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.3.v_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.4.k_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.4.v_cache: torch.Size([1, 8, 313, 128]) -> torch.Size([1, 1, 313, 1024]) - layer.4.output: torch.Size([1, 313, 4096]) -> torch.Size([1, 1, 313, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,940B, BPFP=0.7223 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,216B, BPFP=1.8275 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 51,440B, BPFP=1.2839 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,756B, BPFP=1.9408 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,376B, BPFP=1.4571 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,280B, BPFP=1.9788 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,016B, BPFP=1.4980 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,212B, BPFP=1.9522 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 52,612B, BPFP=1.3132 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,848B, BPFP=1.9930 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 234,040B, BPFP=1.4604 -⌛️ [2/4] FRONTEND: Frontend time: 0.372s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 313, 128]) - layer.0.v_cache: torch.Size([1, 8, 313, 128]) - layer.1.k_cache: torch.Size([1, 8, 313, 128]) - layer.1.v_cache: torch.Size([1, 8, 313, 128]) - layer.2.k_cache: torch.Size([1, 8, 313, 128]) - layer.2.v_cache: torch.Size([1, 8, 313, 128]) - layer.3.k_cache: torch.Size([1, 8, 313, 128]) - layer.3.v_cache: torch.Size([1, 8, 313, 128]) - layer.4.k_cache: torch.Size([1, 8, 313, 128]) - layer.4.v_cache: torch.Size([1, 8, 313, 128]) - layer.4.output: torch.Size([1, 313, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.619s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 313, 128]) - layer.0.v_cache: torch.Size([1, 8, 313, 128]) - layer.1.k_cache: torch.Size([1, 8, 313, 128]) - layer.1.v_cache: torch.Size([1, 8, 313, 128]) - layer.2.k_cache: torch.Size([1, 8, 313, 128]) - layer.2.v_cache: torch.Size([1, 8, 313, 128]) - layer.3.k_cache: torch.Size([1, 8, 313, 128]) - layer.3.v_cache: torch.Size([1, 8, 313, 128]) - layer.4.k_cache: torch.Size([1, 8, 313, 128]) - layer.4.v_cache: torch.Size([1, 8, 313, 128]) - layer.4.output: torch.Size([1, 313, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02366765 6.62005635 - layer.0.v_cache 0.00000027 0.00022940 - layer.1.k_cache 0.00295685 0.59548653 - layer.1.v_cache 0.00000081 0.00079745 - layer.2.k_cache 0.00117427 0.38709878 - layer.2.v_cache 0.00000111 0.00118170 - layer.3.k_cache 0.00131811 0.43016759 - layer.3.v_cache 0.00000214 0.00190414 - layer.4.k_cache 0.00360037 0.75516278 - layer.4.v_cache 0.00000322 0.00333306 - layer.4.output 0.00015721 0.05887470 - ------------------------------------------------------------------------------------- - TOTAL 0.00238240 0.64506547 - (elements=4,487,168) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4487168 -Total Bytes 873736 -BPFP 1.5578 bits/point -EBPFP 3.1155 equivalent bits/point -MSE 0.645065 ----------------------- -------------------------------------------------------- -Time: 1.013s Load: 0.021s, Pack+Encode: 0.372s, Decode+Unpack: 0.619s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 313, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 313, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6451 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample51-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample51-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample52-layer4-item1.zst (59/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample52-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.016s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 329, 128) -Output shape: (1, 329, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.output: torch.Size([1, 329, 4096]) -> torch.Size([1, 1, 329, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,140B, BPFP=0.7157 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,572B, BPFP=1.8658 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,792B, BPFP=1.3011 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,880B, BPFP=2.0156 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,248B, BPFP=1.4782 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,384B, BPFP=2.0513 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,236B, BPFP=1.5254 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,992B, BPFP=2.0182 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,952B, BPFP=1.3286 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,080B, BPFP=2.0678 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,760B, BPFP=1.4768 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.719s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02435147 7.30113522 - layer.0.v_cache 0.00000027 0.00023858 - layer.1.k_cache 0.00282437 0.63176920 - layer.1.v_cache 0.00000081 0.00086408 - layer.2.k_cache 0.00117800 0.39698165 - layer.2.v_cache 0.00000117 0.00125113 - layer.3.k_cache 0.00131145 0.43812603 - layer.3.v_cache 0.00000228 0.00204259 - layer.4.k_cache 0.00354228 0.79064190 - layer.4.v_cache 0.00000348 0.00351852 - layer.4.output 0.00016143 0.06043476 - ------------------------------------------------------------------------------------- - TOTAL 0.00241866 0.70059342 - (elements=4,716,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4716544 -Total Bytes 938036 -BPFP 1.5911 bits/point -EBPFP 3.1821 equivalent bits/point -MSE 0.700593 ----------------------- -------------------------------------------------------- -Time: 1.163s Load: 0.016s, Pack+Encode: 0.429s, Decode+Unpack: 0.719s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7006 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample52-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample52-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample53-layer4-item1.zst (60/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample53-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 330, 128) -Output shape: (1, 330, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.output: torch.Size([1, 330, 4096]) -> torch.Size([1, 1, 330, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,244B, BPFP=0.7160 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,084B, BPFP=1.8723 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,848B, BPFP=1.2985 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,428B, BPFP=2.0224 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,672B, BPFP=1.4837 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,736B, BPFP=2.0534 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,816B, BPFP=1.5345 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,172B, BPFP=2.0164 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,572B, BPFP=1.3393 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,576B, BPFP=2.0733 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 247,824B, BPFP=1.4668 -⌛️ [2/4] FRONTEND: Frontend time: 0.421s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.705s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02327405 6.85349269 - layer.0.v_cache 0.00000026 0.00023649 - layer.1.k_cache 0.00286502 0.62584247 - layer.1.v_cache 0.00000078 0.00085857 - layer.2.k_cache 0.00115867 0.39529206 - layer.2.v_cache 0.00000116 0.00121468 - layer.3.k_cache 0.00132713 0.43561688 - layer.3.v_cache 0.00000212 0.00196699 - layer.4.k_cache 0.00352212 0.78251204 - layer.4.v_cache 0.00000405 0.00352311 - layer.4.output 0.00014449 0.06057661 - ------------------------------------------------------------------------------------- - TOTAL 0.00233809 0.66734732 - (elements=4,730,880) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4730880 -Total Bytes 940972 -BPFP 1.5912 bits/point -EBPFP 3.1824 equivalent bits/point -MSE 0.667347 ----------------------- -------------------------------------------------------- -Time: 1.143s Load: 0.018s, Pack+Encode: 0.421s, Decode+Unpack: 0.705s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6673 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample53-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample53-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample54-layer4-item1.zst (61/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample54-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 320, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 320, 128) -Output shape: (1, 320, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.output: torch.Size([1, 320, 4096]) -> torch.Size([1, 1, 320, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,196B, BPFP=0.6884 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,272B, BPFP=1.7645 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 50,568B, BPFP=1.2346 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,412B, BPFP=1.8899 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 57,580B, BPFP=1.4058 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,044B, BPFP=1.9298 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 59,052B, BPFP=1.4417 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,748B, BPFP=1.8981 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 52,036B, BPFP=1.2704 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,596B, BPFP=1.9433 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 229,428B, BPFP=1.4003 -⌛️ [2/4] FRONTEND: Frontend time: 0.370s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.output: torch.Size([1, 320, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.610s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.output: torch.Size([1, 320, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02461678 6.19005318 - layer.0.v_cache 0.00000026 0.00022876 - layer.1.k_cache 0.00291374 0.57287588 - layer.1.v_cache 0.00000082 0.00083531 - layer.2.k_cache 0.00121384 0.38418100 - layer.2.v_cache 0.00000117 0.00123142 - layer.3.k_cache 0.00132315 0.42387486 - layer.3.v_cache 0.00000219 0.00199526 - layer.4.k_cache 0.00353222 0.73676414 - layer.4.v_cache 0.00000326 0.00343123 - layer.4.output 0.00015660 0.05245805 - ------------------------------------------------------------------------------------- - TOTAL 0.00244527 0.60895023 - (elements=4,587,520) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4587520 -Total Bytes 862932 -BPFP 1.5048 bits/point -EBPFP 3.0097 equivalent bits/point -MSE 0.608950 ----------------------- -------------------------------------------------------- -Time: 0.998s Load: 0.018s, Pack+Encode: 0.370s, Decode+Unpack: 0.610s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 320, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6090 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample54-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample54-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample55-layer4-item1.zst (62/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample55-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 339, 128) -Output shape: (1, 339, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.output: torch.Size([1, 339, 4096]) -> torch.Size([1, 1, 339, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,940B, BPFP=0.7130 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,588B, BPFP=1.8572 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,788B, BPFP=1.3087 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,996B, BPFP=2.0049 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,244B, BPFP=1.4805 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,552B, BPFP=2.0407 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,260B, BPFP=1.5270 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,320B, BPFP=2.0124 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,988B, BPFP=1.3364 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,196B, BPFP=2.0556 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 251,076B, BPFP=1.4466 -⌛️ [2/4] FRONTEND: Frontend time: 0.419s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02391626 6.63021626 - layer.0.v_cache 0.00000026 0.00023266 - layer.1.k_cache 0.00286983 0.63988619 - layer.1.v_cache 0.00000081 0.00086039 - layer.2.k_cache 0.00115809 0.39292561 - layer.2.v_cache 0.00000117 0.00123809 - layer.3.k_cache 0.00131372 0.43231543 - layer.3.v_cache 0.00000240 0.00198107 - layer.4.k_cache 0.00359509 0.76909392 - layer.4.v_cache 0.00000368 0.00347733 - layer.4.output 0.00014345 0.05640498 - ------------------------------------------------------------------------------------- - TOTAL 0.00238822 0.64984620 - (elements=4,859,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4859904 -Total Bytes 959948 -BPFP 1.5802 bits/point -EBPFP 3.1604 equivalent bits/point -MSE 0.649846 ----------------------- -------------------------------------------------------- -Time: 1.144s Load: 0.019s, Pack+Encode: 0.419s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6498 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample55-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample55-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample56-layer4-item1.zst (63/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample56-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 314, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 314, 128) -Output shape: (1, 314, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.0.v_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.1.k_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.1.v_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.2.k_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.2.v_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.3.k_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.3.v_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.4.k_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.4.v_cache: torch.Size([1, 8, 314, 128]) -> torch.Size([1, 1, 314, 1024]) - layer.4.output: torch.Size([1, 314, 4096]) -> torch.Size([1, 1, 314, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,604B, BPFP=0.7117 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 73,184B, BPFP=1.8209 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 51,788B, BPFP=1.2885 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,292B, BPFP=1.9479 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,432B, BPFP=1.4538 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,392B, BPFP=1.9753 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,132B, BPFP=1.4961 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,292B, BPFP=1.9479 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,188B, BPFP=1.3233 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,964B, BPFP=1.9896 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 232,072B, BPFP=1.4435 -⌛️ [2/4] FRONTEND: Frontend time: 0.375s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 314, 128]) - layer.0.v_cache: torch.Size([1, 8, 314, 128]) - layer.1.k_cache: torch.Size([1, 8, 314, 128]) - layer.1.v_cache: torch.Size([1, 8, 314, 128]) - layer.2.k_cache: torch.Size([1, 8, 314, 128]) - layer.2.v_cache: torch.Size([1, 8, 314, 128]) - layer.3.k_cache: torch.Size([1, 8, 314, 128]) - layer.3.v_cache: torch.Size([1, 8, 314, 128]) - layer.4.k_cache: torch.Size([1, 8, 314, 128]) - layer.4.v_cache: torch.Size([1, 8, 314, 128]) - layer.4.output: torch.Size([1, 314, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.598s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 314, 128]) - layer.0.v_cache: torch.Size([1, 8, 314, 128]) - layer.1.k_cache: torch.Size([1, 8, 314, 128]) - layer.1.v_cache: torch.Size([1, 8, 314, 128]) - layer.2.k_cache: torch.Size([1, 8, 314, 128]) - layer.2.v_cache: torch.Size([1, 8, 314, 128]) - layer.3.k_cache: torch.Size([1, 8, 314, 128]) - layer.3.v_cache: torch.Size([1, 8, 314, 128]) - layer.4.k_cache: torch.Size([1, 8, 314, 128]) - layer.4.v_cache: torch.Size([1, 8, 314, 128]) - layer.4.output: torch.Size([1, 314, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02380536 6.57121909 - layer.0.v_cache 0.00000027 0.00024512 - layer.1.k_cache 0.00292133 0.63251316 - layer.1.v_cache 0.00000080 0.00086405 - layer.2.k_cache 0.00120652 0.39696503 - layer.2.v_cache 0.00000113 0.00123966 - layer.3.k_cache 0.00133526 0.42737516 - layer.3.v_cache 0.00000215 0.00196978 - layer.4.k_cache 0.00353655 0.77962110 - layer.4.v_cache 0.00000336 0.00340801 - layer.4.output 0.00016938 0.05579827 - ------------------------------------------------------------------------------------- - TOTAL 0.00239216 0.64561523 - (elements=4,501,504) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4501504 -Total Bytes 873340 -BPFP 1.5521 bits/point -EBPFP 3.1042 equivalent bits/point -MSE 0.645615 ----------------------- -------------------------------------------------------- -Time: 0.991s Load: 0.018s, Pack+Encode: 0.375s, Decode+Unpack: 0.598s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 314, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 314, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6456 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample56-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample56-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample57-layer4-item1.zst (64/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample57-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 344, 128) -Output shape: (1, 344, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.output: torch.Size([1, 344, 4096]) -> torch.Size([1, 1, 344, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,236B, BPFP=0.7094 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,932B, BPFP=1.8607 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,188B, BPFP=1.2988 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,080B, BPFP=2.0004 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,112B, BPFP=1.4787 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,784B, BPFP=2.0391 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,068B, BPFP=1.5232 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,268B, BPFP=2.0046 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,516B, BPFP=1.3289 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,224B, BPFP=2.0491 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 252,876B, BPFP=1.4358 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.709s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02423059 7.04086375 - layer.0.v_cache 0.00000027 0.00023068 - layer.1.k_cache 0.00284490 0.60731129 - layer.1.v_cache 0.00000080 0.00087901 - layer.2.k_cache 0.00115895 0.38692962 - layer.2.v_cache 0.00000119 0.00125701 - layer.3.k_cache 0.00130300 0.42641351 - layer.3.v_cache 0.00000215 0.00200904 - layer.4.k_cache 0.00352154 0.74332140 - layer.4.v_cache 0.00000376 0.00352660 - layer.4.output 0.00013131 0.05033041 - ------------------------------------------------------------------------------------- - TOTAL 0.00239946 0.67243311 - (elements=4,931,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4931584 -Total Bytes 970284 -BPFP 1.5740 bits/point -EBPFP 3.1480 equivalent bits/point -MSE 0.672433 ----------------------- -------------------------------------------------------- -Time: 1.144s Load: 0.019s, Pack+Encode: 0.416s, Decode+Unpack: 0.709s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6724 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample57-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample57-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample58-layer4-item1.zst (65/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample58-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 308, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.015s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 308, 128) -Output shape: (1, 308, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.0.v_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.1.k_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.1.v_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.2.k_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.2.v_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.3.k_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.3.v_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.4.k_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.4.v_cache: torch.Size([1, 8, 308, 128]) -> torch.Size([1, 1, 308, 1024]) - layer.4.output: torch.Size([1, 308, 4096]) -> torch.Size([1, 1, 308, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,544B, BPFP=0.7240 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,688B, BPFP=1.8438 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 50,856B, BPFP=1.2900 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,668B, BPFP=1.9701 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 57,464B, BPFP=1.4576 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 78,932B, BPFP=2.0021 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 59,272B, BPFP=1.5034 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,908B, BPFP=1.9762 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 52,240B, BPFP=1.3251 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,484B, BPFP=2.0161 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 232,276B, BPFP=1.4729 -⌛️ [2/4] FRONTEND: Frontend time: 0.382s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 308, 128]) - layer.0.v_cache: torch.Size([1, 8, 308, 128]) - layer.1.k_cache: torch.Size([1, 8, 308, 128]) - layer.1.v_cache: torch.Size([1, 8, 308, 128]) - layer.2.k_cache: torch.Size([1, 8, 308, 128]) - layer.2.v_cache: torch.Size([1, 8, 308, 128]) - layer.3.k_cache: torch.Size([1, 8, 308, 128]) - layer.3.v_cache: torch.Size([1, 8, 308, 128]) - layer.4.k_cache: torch.Size([1, 8, 308, 128]) - layer.4.v_cache: torch.Size([1, 8, 308, 128]) - layer.4.output: torch.Size([1, 308, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.595s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 308, 128]) - layer.0.v_cache: torch.Size([1, 8, 308, 128]) - layer.1.k_cache: torch.Size([1, 8, 308, 128]) - layer.1.v_cache: torch.Size([1, 8, 308, 128]) - layer.2.k_cache: torch.Size([1, 8, 308, 128]) - layer.2.v_cache: torch.Size([1, 8, 308, 128]) - layer.3.k_cache: torch.Size([1, 8, 308, 128]) - layer.3.v_cache: torch.Size([1, 8, 308, 128]) - layer.4.k_cache: torch.Size([1, 8, 308, 128]) - layer.4.v_cache: torch.Size([1, 8, 308, 128]) - layer.4.output: torch.Size([1, 308, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02345257 6.78822148 - layer.0.v_cache 0.00000027 0.00023328 - layer.1.k_cache 0.00291045 0.65045290 - layer.1.v_cache 0.00000082 0.00086637 - layer.2.k_cache 0.00116968 0.38699442 - layer.2.v_cache 0.00000115 0.00121182 - layer.3.k_cache 0.00131855 0.43701974 - layer.3.v_cache 0.00000224 0.00199521 - layer.4.k_cache 0.00354502 0.75586220 - layer.4.v_cache 0.00000324 0.00341299 - layer.4.output 0.00015550 0.06168847 - ------------------------------------------------------------------------------------- - TOTAL 0.00235900 0.66235888 - (elements=4,415,488) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4415488 -Total Bytes 867332 -BPFP 1.5714 bits/point -EBPFP 3.1429 equivalent bits/point -MSE 0.662359 ----------------------- -------------------------------------------------------- -Time: 0.992s Load: 0.015s, Pack+Encode: 0.382s, Decode+Unpack: 0.595s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 308, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 308, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6624 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample58-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample58-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample59-layer4-item1.zst (66/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample59-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.022s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 339, 128) -Output shape: (1, 339, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) -> torch.Size([1, 1, 339, 1024]) - layer.4.output: torch.Size([1, 339, 4096]) -> torch.Size([1, 1, 339, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,896B, BPFP=0.7120 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,228B, BPFP=1.8489 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,128B, BPFP=1.2935 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,384B, BPFP=1.9908 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,344B, BPFP=1.4598 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,048B, BPFP=2.0291 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,520B, BPFP=1.5100 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,752B, BPFP=1.9993 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,036B, BPFP=1.3144 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,800B, BPFP=2.0465 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,424B, BPFP=1.4601 -⌛️ [2/4] FRONTEND: Frontend time: 0.441s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.716s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 339, 128]) - layer.0.v_cache: torch.Size([1, 8, 339, 128]) - layer.1.k_cache: torch.Size([1, 8, 339, 128]) - layer.1.v_cache: torch.Size([1, 8, 339, 128]) - layer.2.k_cache: torch.Size([1, 8, 339, 128]) - layer.2.v_cache: torch.Size([1, 8, 339, 128]) - layer.3.k_cache: torch.Size([1, 8, 339, 128]) - layer.3.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.k_cache: torch.Size([1, 8, 339, 128]) - layer.4.v_cache: torch.Size([1, 8, 339, 128]) - layer.4.output: torch.Size([1, 339, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02515205 7.30146398 - layer.0.v_cache 0.00000026 0.00023919 - layer.1.k_cache 0.00294413 0.64269182 - layer.1.v_cache 0.00000079 0.00087000 - layer.2.k_cache 0.00117757 0.39250341 - layer.2.v_cache 0.00000116 0.00125358 - layer.3.k_cache 0.00130963 0.43580249 - layer.3.v_cache 0.00000218 0.00205734 - layer.4.k_cache 0.00367572 0.76775160 - layer.4.v_cache 0.00000325 0.00341794 - layer.4.output 0.00014293 0.05565742 - ------------------------------------------------------------------------------------- - TOTAL 0.00248846 0.69790579 - (elements=4,859,904) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4859904 -Total Bytes 956560 -BPFP 1.5746 bits/point -EBPFP 3.1492 equivalent bits/point -MSE 0.697906 ----------------------- -------------------------------------------------------- -Time: 1.179s Load: 0.022s, Pack+Encode: 0.441s, Decode+Unpack: 0.716s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 339, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 339, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6979 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample59-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample59-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample6-layer4-item1.zst (67/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample6-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 355, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.021s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 355, 128) -Output shape: (1, 355, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.0.v_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.1.k_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.1.v_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.2.k_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.2.v_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.3.k_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.3.v_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.4.k_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.4.v_cache: torch.Size([1, 8, 355, 128]) -> torch.Size([1, 1, 355, 1024]) - layer.4.output: torch.Size([1, 355, 4096]) -> torch.Size([1, 1, 355, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,520B, BPFP=0.6937 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 83,576B, BPFP=1.8393 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,792B, BPFP=1.2938 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 90,560B, BPFP=1.9930 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 67,092B, BPFP=1.4765 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 92,004B, BPFP=2.0247 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 68,844B, BPFP=1.5151 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 90,428B, BPFP=1.9901 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 60,412B, BPFP=1.3295 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 92,544B, BPFP=2.0366 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 255,592B, BPFP=1.4062 -⌛️ [2/4] FRONTEND: Frontend time: 0.417s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 355, 128]) - layer.0.v_cache: torch.Size([1, 8, 355, 128]) - layer.1.k_cache: torch.Size([1, 8, 355, 128]) - layer.1.v_cache: torch.Size([1, 8, 355, 128]) - layer.2.k_cache: torch.Size([1, 8, 355, 128]) - layer.2.v_cache: torch.Size([1, 8, 355, 128]) - layer.3.k_cache: torch.Size([1, 8, 355, 128]) - layer.3.v_cache: torch.Size([1, 8, 355, 128]) - layer.4.k_cache: torch.Size([1, 8, 355, 128]) - layer.4.v_cache: torch.Size([1, 8, 355, 128]) - layer.4.output: torch.Size([1, 355, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.702s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 355, 128]) - layer.0.v_cache: torch.Size([1, 8, 355, 128]) - layer.1.k_cache: torch.Size([1, 8, 355, 128]) - layer.1.v_cache: torch.Size([1, 8, 355, 128]) - layer.2.k_cache: torch.Size([1, 8, 355, 128]) - layer.2.v_cache: torch.Size([1, 8, 355, 128]) - layer.3.k_cache: torch.Size([1, 8, 355, 128]) - layer.3.v_cache: torch.Size([1, 8, 355, 128]) - layer.4.k_cache: torch.Size([1, 8, 355, 128]) - layer.4.v_cache: torch.Size([1, 8, 355, 128]) - layer.4.output: torch.Size([1, 355, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02438693 6.73159799 - layer.0.v_cache 0.00000027 0.00023218 - layer.1.k_cache 0.00283684 0.63324241 - layer.1.v_cache 0.00000080 0.00089083 - layer.2.k_cache 0.00116829 0.39302794 - layer.2.v_cache 0.00000120 0.00128196 - layer.3.k_cache 0.00130356 0.43486517 - layer.3.v_cache 0.00000219 0.00198633 - layer.4.k_cache 0.00355311 0.76543115 - layer.4.v_cache 0.00000352 0.00345594 - layer.4.output 0.00013754 0.05198913 - ------------------------------------------------------------------------------------- - TOTAL 0.00241478 0.65528346 - (elements=5,089,280) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5089280 -Total Bytes 991364 -BPFP 1.5584 bits/point -EBPFP 3.1167 equivalent bits/point -MSE 0.655283 ----------------------- -------------------------------------------------------- -Time: 1.140s Load: 0.021s, Pack+Encode: 0.417s, Decode+Unpack: 0.702s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 355, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 355, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6553 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample6-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample6-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample60-layer4-item1.zst (68/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample60-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 334, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 334, 128) -Output shape: (1, 334, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) -> torch.Size([1, 1, 334, 1024]) - layer.4.output: torch.Size([1, 334, 4096]) -> torch.Size([1, 1, 334, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,352B, BPFP=0.7100 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,056B, BPFP=1.8726 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,772B, BPFP=1.3045 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,080B, BPFP=2.0135 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,408B, BPFP=1.4832 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,592B, BPFP=2.0488 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,240B, BPFP=1.5260 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,204B, BPFP=2.0164 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,072B, BPFP=1.3350 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,288B, BPFP=2.0651 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,920B, BPFP=1.4556 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.output: torch.Size([1, 334, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.721s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 334, 128]) - layer.0.v_cache: torch.Size([1, 8, 334, 128]) - layer.1.k_cache: torch.Size([1, 8, 334, 128]) - layer.1.v_cache: torch.Size([1, 8, 334, 128]) - layer.2.k_cache: torch.Size([1, 8, 334, 128]) - layer.2.v_cache: torch.Size([1, 8, 334, 128]) - layer.3.k_cache: torch.Size([1, 8, 334, 128]) - layer.3.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.k_cache: torch.Size([1, 8, 334, 128]) - layer.4.v_cache: torch.Size([1, 8, 334, 128]) - layer.4.output: torch.Size([1, 334, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02444427 6.86909750 - layer.0.v_cache 0.00000026 0.00023624 - layer.1.k_cache 0.00284576 0.63508108 - layer.1.v_cache 0.00000078 0.00086900 - layer.2.k_cache 0.00117801 0.39450548 - layer.2.v_cache 0.00000115 0.00122107 - layer.3.k_cache 0.00132187 0.42851838 - layer.3.v_cache 0.00000218 0.00202002 - layer.4.k_cache 0.00348374 0.75668691 - layer.4.v_cache 0.00000337 0.00351946 - layer.4.output 0.00014793 0.05579316 - ------------------------------------------------------------------------------------- - TOTAL 0.00241951 0.66535198 - (elements=4,788,224) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4788224 -Total Bytes 948984 -BPFP 1.5855 bits/point -EBPFP 3.1711 equivalent bits/point -MSE 0.665352 ----------------------- -------------------------------------------------------- -Time: 1.169s Load: 0.018s, Pack+Encode: 0.429s, Decode+Unpack: 0.721s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 334, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 334, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6654 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample60-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample60-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample61-layer4-item1.zst (69/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample61-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 342, 128) -Output shape: (1, 342, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.output: torch.Size([1, 342, 4096]) -> torch.Size([1, 1, 342, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,272B, BPFP=0.6915 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,032B, BPFP=1.8511 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,140B, BPFP=1.3053 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,548B, BPFP=1.9999 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,840B, BPFP=1.4812 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,916B, BPFP=2.0312 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,520B, BPFP=1.5196 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,484B, BPFP=1.9984 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,004B, BPFP=1.3250 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,464B, BPFP=2.0437 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,020B, BPFP=1.4164 -⌛️ [2/4] FRONTEND: Frontend time: 0.417s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02429651 6.71187337 - layer.0.v_cache 0.00000027 0.00023271 - layer.1.k_cache 0.00293320 0.63086442 - layer.1.v_cache 0.00000079 0.00090200 - layer.2.k_cache 0.00117175 0.40389167 - layer.2.v_cache 0.00000114 0.00126810 - layer.3.k_cache 0.00133100 0.44254486 - layer.3.v_cache 0.00000223 0.00197819 - layer.4.k_cache 0.00354789 0.79567437 - layer.4.v_cache 0.00000344 0.00341788 - layer.4.output 0.00015145 0.05675151 - ------------------------------------------------------------------------------------- - TOTAL 0.00242100 0.65854669 - (elements=4,902,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4902912 -Total Bytes 959240 -BPFP 1.5652 bits/point -EBPFP 3.1304 equivalent bits/point -MSE 0.658547 ----------------------- -------------------------------------------------------- -Time: 1.143s Load: 0.019s, Pack+Encode: 0.417s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6585 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample61-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample61-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample62-layer4-item1.zst (70/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample62-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 312, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 312, 128) -Output shape: (1, 312, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.0.v_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.1.k_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.1.v_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.2.k_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.2.v_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.3.k_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.3.v_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.4.k_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.4.v_cache: torch.Size([1, 8, 312, 128]) -> torch.Size([1, 1, 312, 1024]) - layer.4.output: torch.Size([1, 312, 4096]) -> torch.Size([1, 1, 312, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,144B, BPFP=0.7047 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,420B, BPFP=1.8134 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 51,096B, BPFP=1.2794 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,512B, BPFP=1.9409 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,052B, BPFP=1.4536 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,052B, BPFP=1.9795 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 59,840B, BPFP=1.4984 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,864B, BPFP=1.9497 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 52,132B, BPFP=1.3054 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,324B, BPFP=1.9863 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 230,864B, BPFP=1.4452 -⌛️ [2/4] FRONTEND: Frontend time: 0.357s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 312, 128]) - layer.0.v_cache: torch.Size([1, 8, 312, 128]) - layer.1.k_cache: torch.Size([1, 8, 312, 128]) - layer.1.v_cache: torch.Size([1, 8, 312, 128]) - layer.2.k_cache: torch.Size([1, 8, 312, 128]) - layer.2.v_cache: torch.Size([1, 8, 312, 128]) - layer.3.k_cache: torch.Size([1, 8, 312, 128]) - layer.3.v_cache: torch.Size([1, 8, 312, 128]) - layer.4.k_cache: torch.Size([1, 8, 312, 128]) - layer.4.v_cache: torch.Size([1, 8, 312, 128]) - layer.4.output: torch.Size([1, 312, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.593s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 312, 128]) - layer.0.v_cache: torch.Size([1, 8, 312, 128]) - layer.1.k_cache: torch.Size([1, 8, 312, 128]) - layer.1.v_cache: torch.Size([1, 8, 312, 128]) - layer.2.k_cache: torch.Size([1, 8, 312, 128]) - layer.2.v_cache: torch.Size([1, 8, 312, 128]) - layer.3.k_cache: torch.Size([1, 8, 312, 128]) - layer.3.v_cache: torch.Size([1, 8, 312, 128]) - layer.4.k_cache: torch.Size([1, 8, 312, 128]) - layer.4.v_cache: torch.Size([1, 8, 312, 128]) - layer.4.output: torch.Size([1, 312, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02396220 7.29578810 - layer.0.v_cache 0.00000026 0.00023578 - layer.1.k_cache 0.00299662 0.62905708 - layer.1.v_cache 0.00000081 0.00084773 - layer.2.k_cache 0.00115233 0.40354095 - layer.2.v_cache 0.00000114 0.00126185 - layer.3.k_cache 0.00133293 0.43799669 - layer.3.v_cache 0.00000220 0.00202009 - layer.4.k_cache 0.00362032 0.78024752 - layer.4.v_cache 0.00000325 0.00338490 - layer.4.output 0.00015781 0.05973850 - ------------------------------------------------------------------------------------- - TOTAL 0.00240738 0.69952390 - (elements=4,472,832) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4472832 -Total Bytes 866300 -BPFP 1.5494 bits/point -EBPFP 3.0989 equivalent bits/point -MSE 0.699524 ----------------------- -------------------------------------------------------- -Time: 0.967s Load: 0.017s, Pack+Encode: 0.357s, Decode+Unpack: 0.593s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 312, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 312, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6995 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample62-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample62-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample63-layer4-item1.zst (71/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample63-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 351, 128) -Output shape: (1, 351, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.output: torch.Size([1, 351, 4096]) -> torch.Size([1, 1, 351, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,968B, BPFP=0.6893 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,756B, BPFP=1.8420 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,400B, BPFP=1.2999 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 89,340B, BPFP=1.9885 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 66,204B, BPFP=1.4736 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,664B, BPFP=2.0180 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 68,044B, BPFP=1.5145 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 89,308B, BPFP=1.9878 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,908B, BPFP=1.3334 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 91,220B, BPFP=2.0304 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 256,196B, BPFP=1.4256 -⌛️ [2/4] FRONTEND: Frontend time: 0.429s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.717s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02435836 6.84489280 - layer.0.v_cache 0.00000027 0.00023291 - layer.1.k_cache 0.00294291 0.61450734 - layer.1.v_cache 0.00000079 0.00088099 - layer.2.k_cache 0.00116231 0.39125178 - layer.2.v_cache 0.00000118 0.00123070 - layer.3.k_cache 0.00133045 0.42956208 - layer.3.v_cache 0.00000222 0.00197301 - layer.4.k_cache 0.00357138 0.76107084 - layer.4.v_cache 0.00000341 0.00337439 - layer.4.output 0.00014419 0.05214923 - ------------------------------------------------------------------------------------- - TOTAL 0.00242500 0.66125527 - (elements=5,031,936) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5031936 -Total Bytes 983008 -BPFP 1.5628 bits/point -EBPFP 3.1257 equivalent bits/point -MSE 0.661255 ----------------------- -------------------------------------------------------- -Time: 1.164s Load: 0.018s, Pack+Encode: 0.429s, Decode+Unpack: 0.717s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6613 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample63-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample63-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample64-layer4-item1.zst (72/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample64-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 323, 128) -Output shape: (1, 323, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.output: torch.Size([1, 323, 4096]) -> torch.Size([1, 1, 323, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,200B, BPFP=0.7063 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 76,940B, BPFP=1.8610 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,028B, BPFP=1.3068 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 83,272B, BPFP=2.0141 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 60,896B, BPFP=1.4729 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 84,524B, BPFP=2.0444 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 62,912B, BPFP=1.5217 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 83,068B, BPFP=2.0092 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,156B, BPFP=1.3341 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 85,048B, BPFP=2.0571 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 236,612B, BPFP=1.4308 -⌛️ [2/4] FRONTEND: Frontend time: 0.447s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02343196 7.15182146 - layer.0.v_cache 0.00000026 0.00023491 - layer.1.k_cache 0.00287105 0.65930648 - layer.1.v_cache 0.00000078 0.00086988 - layer.2.k_cache 0.00117289 0.40273897 - layer.2.v_cache 0.00000112 0.00123634 - layer.3.k_cache 0.00130124 0.43786494 - layer.3.v_cache 0.00000213 0.00192257 - layer.4.k_cache 0.00357998 0.79526503 - layer.4.v_cache 0.00000342 0.00336594 - layer.4.output 0.00014651 0.05681843 - ------------------------------------------------------------------------------------- - TOTAL 0.00235363 0.69156430 - (elements=4,630,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4630528 -Total Bytes 911656 -BPFP 1.5750 bits/point -EBPFP 3.1501 equivalent bits/point -MSE 0.691564 ----------------------- -------------------------------------------------------- -Time: 1.172s Load: 0.017s, Pack+Encode: 0.447s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6916 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample64-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample64-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample65-layer4-item1.zst (73/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample65-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 356, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 356, 128) -Output shape: (1, 356, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) -> torch.Size([1, 1, 356, 1024]) - layer.4.output: torch.Size([1, 356, 4096]) -> torch.Size([1, 1, 356, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 32,144B, BPFP=0.7054 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 84,296B, BPFP=1.8499 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 59,064B, BPFP=1.2962 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 90,532B, BPFP=1.9867 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 67,468B, BPFP=1.4806 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 91,884B, BPFP=2.0164 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 69,596B, BPFP=1.5273 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 90,524B, BPFP=1.9866 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 60,228B, BPFP=1.3217 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 92,736B, BPFP=2.0351 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 260,900B, BPFP=1.4314 -⌛️ [2/4] FRONTEND: Frontend time: 0.422s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.output: torch.Size([1, 356, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.711s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 356, 128]) - layer.0.v_cache: torch.Size([1, 8, 356, 128]) - layer.1.k_cache: torch.Size([1, 8, 356, 128]) - layer.1.v_cache: torch.Size([1, 8, 356, 128]) - layer.2.k_cache: torch.Size([1, 8, 356, 128]) - layer.2.v_cache: torch.Size([1, 8, 356, 128]) - layer.3.k_cache: torch.Size([1, 8, 356, 128]) - layer.3.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.k_cache: torch.Size([1, 8, 356, 128]) - layer.4.v_cache: torch.Size([1, 8, 356, 128]) - layer.4.output: torch.Size([1, 356, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02477347 7.11652229 - layer.0.v_cache 0.00000027 0.00024279 - layer.1.k_cache 0.00294478 0.64007461 - layer.1.v_cache 0.00000080 0.00087637 - layer.2.k_cache 0.00115317 0.39309542 - layer.2.v_cache 0.00000113 0.00123156 - layer.3.k_cache 0.00130758 0.44505623 - layer.3.v_cache 0.00000212 0.00198033 - layer.4.k_cache 0.00357902 0.77581899 - layer.4.v_cache 0.00000333 0.00341033 - layer.4.output 0.00014592 0.05718849 - ------------------------------------------------------------------------------------- - TOTAL 0.00245352 0.68621878 - (elements=5,103,616) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5103616 -Total Bytes 999372 -BPFP 1.5665 bits/point -EBPFP 3.1331 equivalent bits/point -MSE 0.686219 ----------------------- -------------------------------------------------------- -Time: 1.152s Load: 0.019s, Pack+Encode: 0.422s, Decode+Unpack: 0.711s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 356, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 356, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6862 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample65-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample65-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample66-layer4-item1.zst (74/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample66-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 342, 128) -Output shape: (1, 342, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) -> torch.Size([1, 1, 342, 1024]) - layer.4.output: torch.Size([1, 342, 4096]) -> torch.Size([1, 1, 342, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,252B, BPFP=0.7139 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,340B, BPFP=1.8581 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,776B, BPFP=1.2970 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,444B, BPFP=1.9975 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,620B, BPFP=1.4762 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,056B, BPFP=2.0344 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,484B, BPFP=1.5187 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,748B, BPFP=2.0045 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,124B, BPFP=1.3278 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,828B, BPFP=2.0520 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 252,768B, BPFP=1.4435 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.699s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 342, 128]) - layer.0.v_cache: torch.Size([1, 8, 342, 128]) - layer.1.k_cache: torch.Size([1, 8, 342, 128]) - layer.1.v_cache: torch.Size([1, 8, 342, 128]) - layer.2.k_cache: torch.Size([1, 8, 342, 128]) - layer.2.v_cache: torch.Size([1, 8, 342, 128]) - layer.3.k_cache: torch.Size([1, 8, 342, 128]) - layer.3.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.k_cache: torch.Size([1, 8, 342, 128]) - layer.4.v_cache: torch.Size([1, 8, 342, 128]) - layer.4.output: torch.Size([1, 342, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02260964 7.07491976 - layer.0.v_cache 0.00000027 0.00023981 - layer.1.k_cache 0.00288942 0.63840409 - layer.1.v_cache 0.00000084 0.00087321 - layer.2.k_cache 0.00116702 0.38802244 - layer.2.v_cache 0.00000122 0.00123424 - layer.3.k_cache 0.00131477 0.43593174 - layer.3.v_cache 0.00000216 0.00195756 - layer.4.k_cache 0.00350203 0.76269620 - layer.4.v_cache 0.00000356 0.00346719 - layer.4.output 0.00013713 0.05677928 - ------------------------------------------------------------------------------------- - TOTAL 0.00228853 0.68106167 - (elements=4,902,912) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4902912 -Total Bytes 965440 -BPFP 1.5753 bits/point -EBPFP 3.1506 equivalent bits/point -MSE 0.681062 ----------------------- -------------------------------------------------------- -Time: 1.134s Load: 0.019s, Pack+Encode: 0.416s, Decode+Unpack: 0.699s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 342, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 342, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6811 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample66-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample66-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample67-layer4-item1.zst (75/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample67-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 360, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.026s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 360, 128) -Output shape: (1, 360, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.0.v_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.1.k_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.1.v_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.2.k_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.2.v_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.3.k_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.3.v_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.4.k_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.4.v_cache: torch.Size([1, 8, 360, 128]) -> torch.Size([1, 1, 360, 1024]) - layer.4.output: torch.Size([1, 360, 4096]) -> torch.Size([1, 1, 360, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 33,096B, BPFP=0.7182 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 85,696B, BPFP=1.8597 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 60,080B, BPFP=1.3038 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 91,924B, BPFP=1.9949 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 68,040B, BPFP=1.4766 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 93,032B, BPFP=2.0189 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 70,244B, BPFP=1.5244 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 91,944B, BPFP=1.9953 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 61,620B, BPFP=1.3372 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 94,280B, BPFP=2.0460 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 267,492B, BPFP=1.4512 -⌛️ [2/4] FRONTEND: Frontend time: 0.436s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 360, 128]) - layer.0.v_cache: torch.Size([1, 8, 360, 128]) - layer.1.k_cache: torch.Size([1, 8, 360, 128]) - layer.1.v_cache: torch.Size([1, 8, 360, 128]) - layer.2.k_cache: torch.Size([1, 8, 360, 128]) - layer.2.v_cache: torch.Size([1, 8, 360, 128]) - layer.3.k_cache: torch.Size([1, 8, 360, 128]) - layer.3.v_cache: torch.Size([1, 8, 360, 128]) - layer.4.k_cache: torch.Size([1, 8, 360, 128]) - layer.4.v_cache: torch.Size([1, 8, 360, 128]) - layer.4.output: torch.Size([1, 360, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.713s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 360, 128]) - layer.0.v_cache: torch.Size([1, 8, 360, 128]) - layer.1.k_cache: torch.Size([1, 8, 360, 128]) - layer.1.v_cache: torch.Size([1, 8, 360, 128]) - layer.2.k_cache: torch.Size([1, 8, 360, 128]) - layer.2.v_cache: torch.Size([1, 8, 360, 128]) - layer.3.k_cache: torch.Size([1, 8, 360, 128]) - layer.3.v_cache: torch.Size([1, 8, 360, 128]) - layer.4.k_cache: torch.Size([1, 8, 360, 128]) - layer.4.v_cache: torch.Size([1, 8, 360, 128]) - layer.4.output: torch.Size([1, 360, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02342221 7.05735541 - layer.0.v_cache 0.00000028 0.00024146 - layer.1.k_cache 0.00285409 0.62205315 - layer.1.v_cache 0.00000079 0.00087369 - layer.2.k_cache 0.00117169 0.39696960 - layer.2.v_cache 0.00000113 0.00123218 - layer.3.k_cache 0.00130395 0.44098795 - layer.3.v_cache 0.00000220 0.00203312 - layer.4.k_cache 0.00353751 0.77585737 - layer.4.v_cache 0.00000341 0.00360504 - layer.4.output 0.00014250 0.05280990 - ------------------------------------------------------------------------------------- - TOTAL 0.00234766 0.67946061 - (elements=5,160,960) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5160960 -Total Bytes 1017448 -BPFP 1.5771 bits/point -EBPFP 3.1543 equivalent bits/point -MSE 0.679461 ----------------------- -------------------------------------------------------- -Time: 1.174s Load: 0.026s, Pack+Encode: 0.436s, Decode+Unpack: 0.713s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 360, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 360, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6795 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample67-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample67-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample68-layer4-item1.zst (76/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample68-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 294, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.016s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 294, 128) -Output shape: (1, 294, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.0.v_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.1.k_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.1.v_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.2.k_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.2.v_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.3.k_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.3.v_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.4.k_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.4.v_cache: torch.Size([1, 8, 294, 128]) -> torch.Size([1, 1, 294, 1024]) - layer.4.output: torch.Size([1, 294, 4096]) -> torch.Size([1, 1, 294, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 27,016B, BPFP=0.7179 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 70,808B, BPFP=1.8816 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 49,360B, BPFP=1.3116 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 76,104B, BPFP=2.0223 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 56,268B, BPFP=1.4952 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 77,416B, BPFP=2.0572 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 57,908B, BPFP=1.5388 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 76,324B, BPFP=2.0282 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 50,616B, BPFP=1.3450 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 78,120B, BPFP=2.0759 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 223,456B, BPFP=1.4845 -⌛️ [2/4] FRONTEND: Frontend time: 0.366s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 294, 128]) - layer.0.v_cache: torch.Size([1, 8, 294, 128]) - layer.1.k_cache: torch.Size([1, 8, 294, 128]) - layer.1.v_cache: torch.Size([1, 8, 294, 128]) - layer.2.k_cache: torch.Size([1, 8, 294, 128]) - layer.2.v_cache: torch.Size([1, 8, 294, 128]) - layer.3.k_cache: torch.Size([1, 8, 294, 128]) - layer.3.v_cache: torch.Size([1, 8, 294, 128]) - layer.4.k_cache: torch.Size([1, 8, 294, 128]) - layer.4.v_cache: torch.Size([1, 8, 294, 128]) - layer.4.output: torch.Size([1, 294, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.606s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 294, 128]) - layer.0.v_cache: torch.Size([1, 8, 294, 128]) - layer.1.k_cache: torch.Size([1, 8, 294, 128]) - layer.1.v_cache: torch.Size([1, 8, 294, 128]) - layer.2.k_cache: torch.Size([1, 8, 294, 128]) - layer.2.v_cache: torch.Size([1, 8, 294, 128]) - layer.3.k_cache: torch.Size([1, 8, 294, 128]) - layer.3.v_cache: torch.Size([1, 8, 294, 128]) - layer.4.k_cache: torch.Size([1, 8, 294, 128]) - layer.4.v_cache: torch.Size([1, 8, 294, 128]) - layer.4.output: torch.Size([1, 294, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02439844 7.21267389 - layer.0.v_cache 0.00000027 0.00023130 - layer.1.k_cache 0.00289989 0.65314556 - layer.1.v_cache 0.00000083 0.00083786 - layer.2.k_cache 0.00119190 0.39427927 - layer.2.v_cache 0.00000115 0.00122203 - layer.3.k_cache 0.00131407 0.43915558 - layer.3.v_cache 0.00000220 0.00196286 - layer.4.k_cache 0.00351532 0.78247833 - layer.4.v_cache 0.00000333 0.00336338 - layer.4.output 0.00015801 0.07132154 - ------------------------------------------------------------------------------------- - TOTAL 0.00242567 0.69818830 - (elements=4,214,784) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4214784 -Total Bytes 843396 -BPFP 1.6008 bits/point -EBPFP 3.2017 equivalent bits/point -MSE 0.698188 ----------------------- -------------------------------------------------------- -Time: 0.988s Load: 0.016s, Pack+Encode: 0.366s, Decode+Unpack: 0.606s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 294, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 294, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6982 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample68-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample68-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample69-layer4-item1.zst (77/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample69-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 338, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 338, 128) -Output shape: (1, 338, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.0.v_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.1.k_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.1.v_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.2.k_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.2.v_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.3.k_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.3.v_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.4.k_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.4.v_cache: torch.Size([1, 8, 338, 128]) -> torch.Size([1, 1, 338, 1024]) - layer.4.output: torch.Size([1, 338, 4096]) -> torch.Size([1, 1, 338, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,960B, BPFP=0.7156 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,720B, BPFP=1.8658 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,424B, BPFP=1.3042 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,772B, BPFP=2.0056 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,088B, BPFP=1.4813 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,512B, BPFP=2.0459 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,148B, BPFP=1.5289 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,332B, BPFP=2.0186 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,528B, BPFP=1.3297 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,352B, BPFP=2.0653 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 252,464B, BPFP=1.4589 -⌛️ [2/4] FRONTEND: Frontend time: 0.415s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 338, 128]) - layer.0.v_cache: torch.Size([1, 8, 338, 128]) - layer.1.k_cache: torch.Size([1, 8, 338, 128]) - layer.1.v_cache: torch.Size([1, 8, 338, 128]) - layer.2.k_cache: torch.Size([1, 8, 338, 128]) - layer.2.v_cache: torch.Size([1, 8, 338, 128]) - layer.3.k_cache: torch.Size([1, 8, 338, 128]) - layer.3.v_cache: torch.Size([1, 8, 338, 128]) - layer.4.k_cache: torch.Size([1, 8, 338, 128]) - layer.4.v_cache: torch.Size([1, 8, 338, 128]) - layer.4.output: torch.Size([1, 338, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.704s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 338, 128]) - layer.0.v_cache: torch.Size([1, 8, 338, 128]) - layer.1.k_cache: torch.Size([1, 8, 338, 128]) - layer.1.v_cache: torch.Size([1, 8, 338, 128]) - layer.2.k_cache: torch.Size([1, 8, 338, 128]) - layer.2.v_cache: torch.Size([1, 8, 338, 128]) - layer.3.k_cache: torch.Size([1, 8, 338, 128]) - layer.3.v_cache: torch.Size([1, 8, 338, 128]) - layer.4.k_cache: torch.Size([1, 8, 338, 128]) - layer.4.v_cache: torch.Size([1, 8, 338, 128]) - layer.4.output: torch.Size([1, 338, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02370679 6.58957144 - layer.0.v_cache 0.00000026 0.00023055 - layer.1.k_cache 0.00288373 0.62356960 - layer.1.v_cache 0.00000083 0.00086349 - layer.2.k_cache 0.00117924 0.39993385 - layer.2.v_cache 0.00000113 0.00123308 - layer.3.k_cache 0.00131114 0.43939719 - layer.3.v_cache 0.00000214 0.00198522 - layer.4.k_cache 0.00355340 0.78068660 - layer.4.v_cache 0.00000355 0.00345123 - layer.4.output 0.00015048 0.05662011 - ------------------------------------------------------------------------------------- - TOTAL 0.00237458 0.64767162 - (elements=4,845,568) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4845568 -Total Bytes 960300 -BPFP 1.5854 bits/point -EBPFP 3.1709 equivalent bits/point -MSE 0.647672 ----------------------- -------------------------------------------------------- -Time: 1.138s Load: 0.019s, Pack+Encode: 0.415s, Decode+Unpack: 0.704s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 338, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 338, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6477 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample69-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample69-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample7-layer4-item1.zst (78/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample7-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 340, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 340, 128) -Output shape: (1, 340, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) -> torch.Size([1, 1, 340, 1024]) - layer.4.output: torch.Size([1, 340, 4096]) -> torch.Size([1, 1, 340, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,576B, BPFP=0.7256 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,968B, BPFP=1.8605 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,708B, BPFP=1.3030 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,156B, BPFP=2.0027 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 64,128B, BPFP=1.4735 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 88,564B, BPFP=2.0350 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 66,244B, BPFP=1.5222 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 87,428B, BPFP=2.0089 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,988B, BPFP=1.3324 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 89,648B, BPFP=2.0599 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 258,080B, BPFP=1.4825 -⌛️ [2/4] FRONTEND: Frontend time: 0.444s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.output: torch.Size([1, 340, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.710s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 340, 128]) - layer.0.v_cache: torch.Size([1, 8, 340, 128]) - layer.1.k_cache: torch.Size([1, 8, 340, 128]) - layer.1.v_cache: torch.Size([1, 8, 340, 128]) - layer.2.k_cache: torch.Size([1, 8, 340, 128]) - layer.2.v_cache: torch.Size([1, 8, 340, 128]) - layer.3.k_cache: torch.Size([1, 8, 340, 128]) - layer.3.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.k_cache: torch.Size([1, 8, 340, 128]) - layer.4.v_cache: torch.Size([1, 8, 340, 128]) - layer.4.output: torch.Size([1, 340, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02367318 7.15567699 - layer.0.v_cache 0.00000026 0.00023211 - layer.1.k_cache 0.00294777 0.63963264 - layer.1.v_cache 0.00000081 0.00087458 - layer.2.k_cache 0.00121079 0.39186608 - layer.2.v_cache 0.00000125 0.00122800 - layer.3.k_cache 0.00131075 0.43632449 - layer.3.v_cache 0.00000218 0.00200799 - layer.4.k_cache 0.00357831 0.75340801 - layer.4.v_cache 0.00000337 0.00346024 - layer.4.output 0.00014951 0.06346188 - ------------------------------------------------------------------------------------- - TOTAL 0.00238048 0.68846847 - (elements=4,874,240) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4874240 -Total Bytes 968488 -BPFP 1.5896 bits/point -EBPFP 3.1791 equivalent bits/point -MSE 0.688468 ----------------------- -------------------------------------------------------- -Time: 1.171s Load: 0.017s, Pack+Encode: 0.444s, Decode+Unpack: 0.710s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 340, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 340, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6885 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample7-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample7-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample70-layer4-item1.zst (79/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample70-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 320, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 320, 128) -Output shape: (1, 320, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) -> torch.Size([1, 1, 320, 1024]) - layer.4.output: torch.Size([1, 320, 4096]) -> torch.Size([1, 1, 320, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,368B, BPFP=0.6926 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,340B, BPFP=1.7661 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 50,872B, BPFP=1.2420 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,752B, BPFP=1.8982 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 57,648B, BPFP=1.4074 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,008B, BPFP=1.9289 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 59,108B, BPFP=1.4431 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,744B, BPFP=1.8980 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 52,120B, BPFP=1.2725 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 79,416B, BPFP=1.9389 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 226,848B, BPFP=1.3846 -⌛️ [2/4] FRONTEND: Frontend time: 0.370s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.output: torch.Size([1, 320, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.594s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 320, 128]) - layer.0.v_cache: torch.Size([1, 8, 320, 128]) - layer.1.k_cache: torch.Size([1, 8, 320, 128]) - layer.1.v_cache: torch.Size([1, 8, 320, 128]) - layer.2.k_cache: torch.Size([1, 8, 320, 128]) - layer.2.v_cache: torch.Size([1, 8, 320, 128]) - layer.3.k_cache: torch.Size([1, 8, 320, 128]) - layer.3.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.k_cache: torch.Size([1, 8, 320, 128]) - layer.4.v_cache: torch.Size([1, 8, 320, 128]) - layer.4.output: torch.Size([1, 320, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02419060 6.11682205 - layer.0.v_cache 0.00000027 0.00022387 - layer.1.k_cache 0.00288080 0.56005840 - layer.1.v_cache 0.00000084 0.00086193 - layer.2.k_cache 0.00117023 0.37223935 - layer.2.v_cache 0.00000114 0.00119791 - layer.3.k_cache 0.00131996 0.41136651 - layer.3.v_cache 0.00000215 0.00188867 - layer.4.k_cache 0.00348745 0.73165822 - layer.4.v_cache 0.00000346 0.00330597 - layer.4.output 0.00014756 0.04635431 - ------------------------------------------------------------------------------------- - TOTAL 0.00240337 0.59893144 - (elements=4,587,520) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4587520 -Total Bytes 861224 -BPFP 1.5019 bits/point -EBPFP 3.0037 equivalent bits/point -MSE 0.598931 ----------------------- -------------------------------------------------------- -Time: 0.982s Load: 0.018s, Pack+Encode: 0.370s, Decode+Unpack: 0.594s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 320, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 320, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.5989 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample70-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample70-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample71-layer4-item1.zst (80/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample71-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 328, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 328, 128) -Output shape: (1, 328, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.output: torch.Size([1, 328, 4096]) -> torch.Size([1, 1, 328, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,556B, BPFP=0.7278 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,524B, BPFP=1.8703 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,516B, BPFP=1.2985 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,208B, BPFP=2.0057 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,684B, BPFP=1.4692 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,592B, BPFP=2.0387 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,476B, BPFP=1.5357 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,380B, BPFP=2.0098 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,940B, BPFP=1.3324 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,460B, BPFP=2.0594 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 245,916B, BPFP=1.4643 -⌛️ [2/4] FRONTEND: Frontend time: 0.421s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.output: torch.Size([1, 328, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.707s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.output: torch.Size([1, 328, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02411815 7.72713415 - layer.0.v_cache 0.00000026 0.00023830 - layer.1.k_cache 0.00293126 0.63985611 - layer.1.v_cache 0.00000079 0.00085341 - layer.2.k_cache 0.00115822 0.38859414 - layer.2.v_cache 0.00000116 0.00120417 - layer.3.k_cache 0.00132993 0.43825926 - layer.3.v_cache 0.00000214 0.00194721 - layer.4.k_cache 0.00358323 0.77968276 - layer.4.v_cache 0.00000336 0.00336699 - layer.4.output 0.00014415 0.05456305 - ------------------------------------------------------------------------------------- - TOTAL 0.00240751 0.72852776 - (elements=4,702,208) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4702208 -Total Bytes 932252 -BPFP 1.5861 bits/point -EBPFP 3.1721 equivalent bits/point -MSE 0.728528 ----------------------- -------------------------------------------------------- -Time: 1.145s Load: 0.018s, Pack+Encode: 0.421s, Decode+Unpack: 0.707s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 328, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7285 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample71-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample71-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample72-layer4-item1.zst (81/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample72-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 328, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 328, 128) -Output shape: (1, 328, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) -> torch.Size([1, 1, 328, 1024]) - layer.4.output: torch.Size([1, 328, 4096]) -> torch.Size([1, 1, 328, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,488B, BPFP=0.7024 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,200B, BPFP=1.8626 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,680B, BPFP=1.3024 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,200B, BPFP=2.0055 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,948B, BPFP=1.4755 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,052B, BPFP=2.0496 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,416B, BPFP=1.5343 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,592B, BPFP=2.0149 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,968B, BPFP=1.3331 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,144B, BPFP=2.0518 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 243,960B, BPFP=1.4527 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.output: torch.Size([1, 328, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.705s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 328, 128]) - layer.0.v_cache: torch.Size([1, 8, 328, 128]) - layer.1.k_cache: torch.Size([1, 8, 328, 128]) - layer.1.v_cache: torch.Size([1, 8, 328, 128]) - layer.2.k_cache: torch.Size([1, 8, 328, 128]) - layer.2.v_cache: torch.Size([1, 8, 328, 128]) - layer.3.k_cache: torch.Size([1, 8, 328, 128]) - layer.3.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.k_cache: torch.Size([1, 8, 328, 128]) - layer.4.v_cache: torch.Size([1, 8, 328, 128]) - layer.4.output: torch.Size([1, 328, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02582944 7.35407313 - layer.0.v_cache 0.00000026 0.00023484 - layer.1.k_cache 0.00293683 0.64545357 - layer.1.v_cache 0.00000081 0.00087788 - layer.2.k_cache 0.00118071 0.39756231 - layer.2.v_cache 0.00000117 0.00124387 - layer.3.k_cache 0.00133336 0.43343577 - layer.3.v_cache 0.00000214 0.00191767 - layer.4.k_cache 0.00361319 0.77175457 - layer.4.v_cache 0.00000321 0.00330425 - layer.4.output 0.00015382 0.05705876 - ------------------------------------------------------------------------------------- - TOTAL 0.00253689 0.70272092 - (elements=4,702,208) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4702208 -Total Bytes 929648 -BPFP 1.5816 bits/point -EBPFP 3.1633 equivalent bits/point -MSE 0.702721 ----------------------- -------------------------------------------------------- -Time: 1.139s Load: 0.018s, Pack+Encode: 0.416s, Decode+Unpack: 0.705s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 328, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 328, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7027 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample72-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample72-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample73-layer4-item1.zst (82/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample73-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 352, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 352, 128) -Output shape: (1, 352, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.0.v_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.1.k_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.1.v_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.2.k_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.2.v_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.3.k_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.3.v_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.4.k_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.4.v_cache: torch.Size([1, 8, 352, 128]) -> torch.Size([1, 1, 352, 1024]) - layer.4.output: torch.Size([1, 352, 4096]) -> torch.Size([1, 1, 352, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,476B, BPFP=0.6986 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,608B, BPFP=1.8335 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,196B, BPFP=1.2916 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,896B, BPFP=1.9730 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,900B, BPFP=1.4626 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,704B, BPFP=2.0131 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 68,264B, BPFP=1.5151 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 89,536B, BPFP=1.9872 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,504B, BPFP=1.3207 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 91,664B, BPFP=2.0344 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 257,936B, BPFP=1.4312 -⌛️ [2/4] FRONTEND: Frontend time: 0.422s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 352, 128]) - layer.0.v_cache: torch.Size([1, 8, 352, 128]) - layer.1.k_cache: torch.Size([1, 8, 352, 128]) - layer.1.v_cache: torch.Size([1, 8, 352, 128]) - layer.2.k_cache: torch.Size([1, 8, 352, 128]) - layer.2.v_cache: torch.Size([1, 8, 352, 128]) - layer.3.k_cache: torch.Size([1, 8, 352, 128]) - layer.3.v_cache: torch.Size([1, 8, 352, 128]) - layer.4.k_cache: torch.Size([1, 8, 352, 128]) - layer.4.v_cache: torch.Size([1, 8, 352, 128]) - layer.4.output: torch.Size([1, 352, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.706s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 352, 128]) - layer.0.v_cache: torch.Size([1, 8, 352, 128]) - layer.1.k_cache: torch.Size([1, 8, 352, 128]) - layer.1.v_cache: torch.Size([1, 8, 352, 128]) - layer.2.k_cache: torch.Size([1, 8, 352, 128]) - layer.2.v_cache: torch.Size([1, 8, 352, 128]) - layer.3.k_cache: torch.Size([1, 8, 352, 128]) - layer.3.v_cache: torch.Size([1, 8, 352, 128]) - layer.4.k_cache: torch.Size([1, 8, 352, 128]) - layer.4.v_cache: torch.Size([1, 8, 352, 128]) - layer.4.output: torch.Size([1, 352, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02395736 6.81430886 - layer.0.v_cache 0.00000026 0.00023729 - layer.1.k_cache 0.00288394 0.63785341 - layer.1.v_cache 0.00000077 0.00087295 - layer.2.k_cache 0.00114480 0.39111094 - layer.2.v_cache 0.00000114 0.00125890 - layer.3.k_cache 0.00130173 0.43320339 - layer.3.v_cache 0.00000215 0.00205092 - layer.4.k_cache 0.00352615 0.76576216 - layer.4.v_cache 0.00000336 0.00362866 - layer.4.output 0.00014232 0.05325048 - ------------------------------------------------------------------------------------- - TOTAL 0.00238507 0.66166353 - (elements=5,046,272) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5046272 -Total Bytes 984684 -BPFP 1.5610 bits/point -EBPFP 3.1221 equivalent bits/point -MSE 0.661664 ----------------------- -------------------------------------------------------- -Time: 1.147s Load: 0.018s, Pack+Encode: 0.422s, Decode+Unpack: 0.706s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 352, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 352, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6617 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample73-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample73-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample74-layer4-item1.zst (83/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample74-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.016s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 323, 128) -Output shape: (1, 323, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.output: torch.Size([1, 323, 4096]) -> torch.Size([1, 1, 323, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,700B, BPFP=0.7426 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 77,944B, BPFP=1.8853 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,072B, BPFP=1.3079 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 83,208B, BPFP=2.0126 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,492B, BPFP=1.4873 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,192B, BPFP=2.0606 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 63,408B, BPFP=1.5337 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,060B, BPFP=2.0332 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,464B, BPFP=1.3415 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 85,792B, BPFP=2.0751 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 246,404B, BPFP=1.4900 -⌛️ [2/4] FRONTEND: Frontend time: 0.413s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.708s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02435790 6.87534618 - layer.0.v_cache 0.00000027 0.00023237 - layer.1.k_cache 0.00290123 0.63433451 - layer.1.v_cache 0.00000082 0.00085439 - layer.2.k_cache 0.00116542 0.40272040 - layer.2.v_cache 0.00000118 0.00126060 - layer.3.k_cache 0.00132756 0.43235609 - layer.3.v_cache 0.00000224 0.00203235 - layer.4.k_cache 0.00351721 0.76173812 - layer.4.v_cache 0.00000335 0.00349675 - layer.4.output 0.00015064 0.06053721 - ------------------------------------------------------------------------------------- - TOTAL 0.00241998 0.66832290 - (elements=4,630,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4630528 -Total Bytes 927736 -BPFP 1.6028 bits/point -EBPFP 3.2056 equivalent bits/point -MSE 0.668323 ----------------------- -------------------------------------------------------- -Time: 1.137s Load: 0.016s, Pack+Encode: 0.413s, Decode+Unpack: 0.708s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6683 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample74-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample74-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample75-layer4-item1.zst (84/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample75-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 327, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 327, 128) -Output shape: (1, 327, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) -> torch.Size([1, 1, 327, 1024]) - layer.4.output: torch.Size([1, 327, 4096]) -> torch.Size([1, 1, 327, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,672B, BPFP=0.7328 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,324B, BPFP=1.8713 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,960B, BPFP=1.3131 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,348B, BPFP=2.0152 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,008B, BPFP=1.4815 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,968B, BPFP=2.0539 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,284B, BPFP=1.5358 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,556B, BPFP=2.0202 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,800B, BPFP=1.3331 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,356B, BPFP=2.0632 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 247,688B, BPFP=1.4794 -⌛️ [2/4] FRONTEND: Frontend time: 0.426s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.output: torch.Size([1, 327, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.719s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 327, 128]) - layer.0.v_cache: torch.Size([1, 8, 327, 128]) - layer.1.k_cache: torch.Size([1, 8, 327, 128]) - layer.1.v_cache: torch.Size([1, 8, 327, 128]) - layer.2.k_cache: torch.Size([1, 8, 327, 128]) - layer.2.v_cache: torch.Size([1, 8, 327, 128]) - layer.3.k_cache: torch.Size([1, 8, 327, 128]) - layer.3.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.k_cache: torch.Size([1, 8, 327, 128]) - layer.4.v_cache: torch.Size([1, 8, 327, 128]) - layer.4.output: torch.Size([1, 327, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02420299 7.27135894 - layer.0.v_cache 0.00000027 0.00023216 - layer.1.k_cache 0.00291079 0.63877836 - layer.1.v_cache 0.00000081 0.00087829 - layer.2.k_cache 0.00118771 0.39133983 - layer.2.v_cache 0.00000120 0.00124679 - layer.3.k_cache 0.00132021 0.43342474 - layer.3.v_cache 0.00000219 0.00199344 - layer.4.k_cache 0.00351552 0.76696315 - layer.4.v_cache 0.00000327 0.00338544 - layer.4.output 0.00014986 0.05438827 - ------------------------------------------------------------------------------------- - TOTAL 0.00241031 0.69479673 - (elements=4,687,872) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4687872 -Total Bytes 934964 -BPFP 1.5955 bits/point -EBPFP 3.1911 equivalent bits/point -MSE 0.694797 ----------------------- -------------------------------------------------------- -Time: 1.162s Load: 0.018s, Pack+Encode: 0.426s, Decode+Unpack: 0.719s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 327, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 327, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6948 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample75-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample75-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample76-layer4-item1.zst (85/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample76-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 330, 128) -Output shape: (1, 330, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) -> torch.Size([1, 1, 330, 1024]) - layer.4.output: torch.Size([1, 330, 4096]) -> torch.Size([1, 1, 330, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,188B, BPFP=0.7147 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,916B, BPFP=1.8683 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,124B, BPFP=1.3050 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,576B, BPFP=2.0023 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,368B, BPFP=1.4765 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,272B, BPFP=2.0424 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,668B, BPFP=1.5310 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,000B, BPFP=2.0123 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,156B, BPFP=1.3295 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,924B, BPFP=2.0579 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 248,852B, BPFP=1.4728 -⌛️ [2/4] FRONTEND: Frontend time: 0.430s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.708s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 330, 128]) - layer.0.v_cache: torch.Size([1, 8, 330, 128]) - layer.1.k_cache: torch.Size([1, 8, 330, 128]) - layer.1.v_cache: torch.Size([1, 8, 330, 128]) - layer.2.k_cache: torch.Size([1, 8, 330, 128]) - layer.2.v_cache: torch.Size([1, 8, 330, 128]) - layer.3.k_cache: torch.Size([1, 8, 330, 128]) - layer.3.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.k_cache: torch.Size([1, 8, 330, 128]) - layer.4.v_cache: torch.Size([1, 8, 330, 128]) - layer.4.output: torch.Size([1, 330, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02537568 6.98574663 - layer.0.v_cache 0.00000027 0.00022748 - layer.1.k_cache 0.00294190 0.63775773 - layer.1.v_cache 0.00000076 0.00080771 - layer.2.k_cache 0.00116180 0.39267296 - layer.2.v_cache 0.00000113 0.00116265 - layer.3.k_cache 0.00131353 0.43051957 - layer.3.v_cache 0.00000217 0.00188204 - layer.4.k_cache 0.00362690 0.78161945 - layer.4.v_cache 0.00000322 0.00327143 - layer.4.output 0.00015895 0.05971904 - ------------------------------------------------------------------------------------- - TOTAL 0.00250451 0.67675313 - (elements=4,730,880) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4730880 -Total Bytes 939044 -BPFP 1.5879 bits/point -EBPFP 3.1759 equivalent bits/point -MSE 0.676753 ----------------------- -------------------------------------------------------- -Time: 1.156s Load: 0.018s, Pack+Encode: 0.430s, Decode+Unpack: 0.708s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 330, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 330, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6768 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample76-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample76-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample77-layer4-item1.zst (86/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample77-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 329, 128) -Output shape: (1, 329, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.output: torch.Size([1, 329, 4096]) -> torch.Size([1, 1, 329, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,360B, BPFP=0.6972 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,444B, BPFP=1.8627 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,944B, BPFP=1.3047 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,892B, BPFP=2.0159 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,032B, BPFP=1.4730 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,076B, BPFP=2.0440 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,388B, BPFP=1.5290 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 84,532B, BPFP=2.0073 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,812B, BPFP=1.3253 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,532B, BPFP=2.0548 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 243,504B, BPFP=1.4456 -⌛️ [2/4] FRONTEND: Frontend time: 0.438s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.708s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02321163 7.30994728 - layer.0.v_cache 0.00000026 0.00023649 - layer.1.k_cache 0.00292509 0.63890475 - layer.1.v_cache 0.00000078 0.00085574 - layer.2.k_cache 0.00118155 0.39546644 - layer.2.v_cache 0.00000112 0.00119594 - layer.3.k_cache 0.00133849 0.43915815 - layer.3.v_cache 0.00000208 0.00188412 - layer.4.k_cache 0.00360187 0.78658946 - layer.4.v_cache 0.00000304 0.00324297 - layer.4.output 0.00015018 0.05890135 - ------------------------------------------------------------------------------------- - TOTAL 0.00234762 0.70093477 - (elements=4,716,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4716544 -Total Bytes 930516 -BPFP 1.5783 bits/point -EBPFP 3.1566 equivalent bits/point -MSE 0.700935 ----------------------- -------------------------------------------------------- -Time: 1.164s Load: 0.018s, Pack+Encode: 0.438s, Decode+Unpack: 0.708s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7009 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample77-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample77-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample78-layer4-item1.zst (87/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample78-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 333, 128) -Output shape: (1, 333, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.output: torch.Size([1, 333, 4096]) -> torch.Size([1, 1, 333, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,316B, BPFP=0.7112 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 79,236B, BPFP=1.8590 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,548B, BPFP=1.3032 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,332B, BPFP=2.0020 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,084B, BPFP=1.4800 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,904B, BPFP=2.0389 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,316B, BPFP=1.5324 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,900B, BPFP=2.0153 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,788B, BPFP=1.3323 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,712B, BPFP=2.0578 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 247,392B, BPFP=1.4510 -⌛️ [2/4] FRONTEND: Frontend time: 0.412s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.695s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02306361 6.86613249 - layer.0.v_cache 0.00000026 0.00023265 - layer.1.k_cache 0.00286224 0.63532007 - layer.1.v_cache 0.00000078 0.00083567 - layer.2.k_cache 0.00116467 0.39949434 - layer.2.v_cache 0.00000115 0.00121384 - layer.3.k_cache 0.00135416 0.44441493 - layer.3.v_cache 0.00000234 0.00201365 - layer.4.k_cache 0.00356469 0.78312257 - layer.4.v_cache 0.00000325 0.00339660 - layer.4.output 0.00015110 0.05988624 - ------------------------------------------------------------------------------------- - TOTAL 0.00233011 0.66969441 - (elements=4,773,888) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4773888 -Total Bytes 943528 -BPFP 1.5811 bits/point -EBPFP 3.1623 equivalent bits/point -MSE 0.669694 ----------------------- -------------------------------------------------------- -Time: 1.126s Load: 0.019s, Pack+Encode: 0.412s, Decode+Unpack: 0.695s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6697 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample78-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample78-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample79-layer4-item1.zst (88/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample79-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 304, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 304, 128) -Output shape: (1, 304, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.0.v_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.1.k_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.1.v_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.2.k_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.2.v_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.3.k_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.3.v_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.4.k_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.4.v_cache: torch.Size([1, 8, 304, 128]) -> torch.Size([1, 1, 304, 1024]) - layer.4.output: torch.Size([1, 304, 4096]) -> torch.Size([1, 1, 304, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 27,432B, BPFP=0.7050 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 71,796B, BPFP=1.8451 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 50,060B, BPFP=1.2865 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 77,048B, BPFP=1.9801 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 56,960B, BPFP=1.4638 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 78,352B, BPFP=2.0136 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 58,524B, BPFP=1.5040 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 77,100B, BPFP=1.9814 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 51,116B, BPFP=1.3136 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 78,800B, BPFP=2.0251 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 226,088B, BPFP=1.4526 -⌛️ [2/4] FRONTEND: Frontend time: 0.366s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 304, 128]) - layer.0.v_cache: torch.Size([1, 8, 304, 128]) - layer.1.k_cache: torch.Size([1, 8, 304, 128]) - layer.1.v_cache: torch.Size([1, 8, 304, 128]) - layer.2.k_cache: torch.Size([1, 8, 304, 128]) - layer.2.v_cache: torch.Size([1, 8, 304, 128]) - layer.3.k_cache: torch.Size([1, 8, 304, 128]) - layer.3.v_cache: torch.Size([1, 8, 304, 128]) - layer.4.k_cache: torch.Size([1, 8, 304, 128]) - layer.4.v_cache: torch.Size([1, 8, 304, 128]) - layer.4.output: torch.Size([1, 304, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.591s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 304, 128]) - layer.0.v_cache: torch.Size([1, 8, 304, 128]) - layer.1.k_cache: torch.Size([1, 8, 304, 128]) - layer.1.v_cache: torch.Size([1, 8, 304, 128]) - layer.2.k_cache: torch.Size([1, 8, 304, 128]) - layer.2.v_cache: torch.Size([1, 8, 304, 128]) - layer.3.k_cache: torch.Size([1, 8, 304, 128]) - layer.3.v_cache: torch.Size([1, 8, 304, 128]) - layer.4.k_cache: torch.Size([1, 8, 304, 128]) - layer.4.v_cache: torch.Size([1, 8, 304, 128]) - layer.4.output: torch.Size([1, 304, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02416075 7.37198197 - layer.0.v_cache 0.00000027 0.00023164 - layer.1.k_cache 0.00290015 0.64365000 - layer.1.v_cache 0.00000081 0.00085865 - layer.2.k_cache 0.00116609 0.39156791 - layer.2.v_cache 0.00000115 0.00122446 - layer.3.k_cache 0.00131777 0.42682457 - layer.3.v_cache 0.00000216 0.00197026 - layer.4.k_cache 0.00349415 0.74312190 - layer.4.v_cache 0.00000336 0.00342233 - layer.4.output 0.00015143 0.05579061 - ------------------------------------------------------------------------------------- - TOTAL 0.00240374 0.70057258 - (elements=4,358,144) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4358144 -Total Bytes 853276 -BPFP 1.5663 bits/point -EBPFP 3.1326 equivalent bits/point -MSE 0.700573 ----------------------- -------------------------------------------------------- -Time: 0.976s Load: 0.019s, Pack+Encode: 0.366s, Decode+Unpack: 0.591s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 304, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 304, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.7006 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample79-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample79-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample8-layer4-item1.zst (89/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample8-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.019s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 351, 128) -Output shape: (1, 351, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) -> torch.Size([1, 1, 351, 1024]) - layer.4.output: torch.Size([1, 351, 4096]) -> torch.Size([1, 1, 351, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,200B, BPFP=0.6944 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 82,276B, BPFP=1.8313 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 58,120B, BPFP=1.2936 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,532B, BPFP=1.9705 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,776B, BPFP=1.4640 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,472B, BPFP=2.0137 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 68,232B, BPFP=1.5187 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 89,288B, BPFP=1.9874 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,816B, BPFP=1.3314 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 91,596B, BPFP=2.0387 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 258,636B, BPFP=1.4392 -⌛️ [2/4] FRONTEND: Frontend time: 0.418s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.730s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 351, 128]) - layer.0.v_cache: torch.Size([1, 8, 351, 128]) - layer.1.k_cache: torch.Size([1, 8, 351, 128]) - layer.1.v_cache: torch.Size([1, 8, 351, 128]) - layer.2.k_cache: torch.Size([1, 8, 351, 128]) - layer.2.v_cache: torch.Size([1, 8, 351, 128]) - layer.3.k_cache: torch.Size([1, 8, 351, 128]) - layer.3.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.k_cache: torch.Size([1, 8, 351, 128]) - layer.4.v_cache: torch.Size([1, 8, 351, 128]) - layer.4.output: torch.Size([1, 351, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02287757 7.10465286 - layer.0.v_cache 0.00000026 0.00022505 - layer.1.k_cache 0.00288319 0.63065910 - layer.1.v_cache 0.00000079 0.00082948 - layer.2.k_cache 0.00116994 0.39080841 - layer.2.v_cache 0.00000119 0.00123793 - layer.3.k_cache 0.00131288 0.43219438 - layer.3.v_cache 0.00000224 0.00199477 - layer.4.k_cache 0.00358885 0.76978226 - layer.4.v_cache 0.00000339 0.00337952 - layer.4.output 0.00014945 0.05533618 - ------------------------------------------------------------------------------------- - TOTAL 0.00231701 0.68265061 - (elements=5,031,936) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5031936 -Total Bytes 983944 -BPFP 1.5643 bits/point -EBPFP 3.1286 equivalent bits/point -MSE 0.682651 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.019s, Pack+Encode: 0.418s, Decode+Unpack: 0.730s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 351, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 351, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6827 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample8-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample8-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample80-layer4-item1.zst (90/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample80-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 333, 128) -Output shape: (1, 333, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) -> torch.Size([1, 1, 333, 1024]) - layer.4.output: torch.Size([1, 333, 4096]) -> torch.Size([1, 1, 333, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,452B, BPFP=0.6910 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,476B, BPFP=1.8411 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 55,416B, BPFP=1.3001 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,932B, BPFP=1.9926 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,764B, BPFP=1.4725 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,884B, BPFP=2.0384 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,884B, BPFP=1.5222 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,552B, BPFP=2.0071 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,320B, BPFP=1.3213 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,420B, BPFP=2.0510 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 244,560B, BPFP=1.4344 -⌛️ [2/4] FRONTEND: Frontend time: 0.437s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.708s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 333, 128]) - layer.0.v_cache: torch.Size([1, 8, 333, 128]) - layer.1.k_cache: torch.Size([1, 8, 333, 128]) - layer.1.v_cache: torch.Size([1, 8, 333, 128]) - layer.2.k_cache: torch.Size([1, 8, 333, 128]) - layer.2.v_cache: torch.Size([1, 8, 333, 128]) - layer.3.k_cache: torch.Size([1, 8, 333, 128]) - layer.3.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.k_cache: torch.Size([1, 8, 333, 128]) - layer.4.v_cache: torch.Size([1, 8, 333, 128]) - layer.4.output: torch.Size([1, 333, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02328842 7.05970228 - layer.0.v_cache 0.00000026 0.00023770 - layer.1.k_cache 0.00290993 0.64801167 - layer.1.v_cache 0.00000078 0.00084506 - layer.2.k_cache 0.00115179 0.39611807 - layer.2.v_cache 0.00000114 0.00123501 - layer.3.k_cache 0.00134347 0.44087526 - layer.3.v_cache 0.00000212 0.00199950 - layer.4.k_cache 0.00362086 0.78681593 - layer.4.v_cache 0.00000365 0.00332393 - layer.4.output 0.00013924 0.05539243 - ------------------------------------------------------------------------------------- - TOTAL 0.00234853 0.68290958 - (elements=4,773,888) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4773888 -Total Bytes 936660 -BPFP 1.5696 bits/point -EBPFP 3.1393 equivalent bits/point -MSE 0.682910 ----------------------- -------------------------------------------------------- -Time: 1.163s Load: 0.018s, Pack+Encode: 0.437s, Decode+Unpack: 0.708s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 333, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 333, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6829 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample80-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample80-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample81-layer4-item1.zst (91/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample81-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 319, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 319, 128) -Output shape: (1, 319, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) -> torch.Size([1, 1, 319, 1024]) - layer.4.output: torch.Size([1, 319, 4096]) -> torch.Size([1, 1, 319, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 28,728B, BPFP=0.7036 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 72,876B, BPFP=1.7848 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 52,080B, BPFP=1.2755 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 78,232B, BPFP=1.9159 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 58,848B, BPFP=1.4412 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 79,828B, BPFP=1.9550 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 60,292B, BPFP=1.4766 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 78,560B, BPFP=1.9240 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 53,628B, BPFP=1.3134 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 80,324B, BPFP=1.9672 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 234,644B, BPFP=1.4366 -⌛️ [2/4] FRONTEND: Frontend time: 0.365s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.output: torch.Size([1, 319, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.595s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 319, 128]) - layer.0.v_cache: torch.Size([1, 8, 319, 128]) - layer.1.k_cache: torch.Size([1, 8, 319, 128]) - layer.1.v_cache: torch.Size([1, 8, 319, 128]) - layer.2.k_cache: torch.Size([1, 8, 319, 128]) - layer.2.v_cache: torch.Size([1, 8, 319, 128]) - layer.3.k_cache: torch.Size([1, 8, 319, 128]) - layer.3.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.k_cache: torch.Size([1, 8, 319, 128]) - layer.4.v_cache: torch.Size([1, 8, 319, 128]) - layer.4.output: torch.Size([1, 319, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02501136 6.38373664 - layer.0.v_cache 0.00000027 0.00022333 - layer.1.k_cache 0.00284013 0.58100470 - layer.1.v_cache 0.00000084 0.00084195 - layer.2.k_cache 0.00117936 0.38968386 - layer.2.v_cache 0.00000120 0.00124656 - layer.3.k_cache 0.00130692 0.41988710 - layer.3.v_cache 0.00000226 0.00196924 - layer.4.k_cache 0.00361670 0.75668856 - layer.4.v_cache 0.00000371 0.00342061 - layer.4.output 0.00016871 0.05493789 - ------------------------------------------------------------------------------------- - TOTAL 0.00247411 0.62560387 - (elements=4,573,184) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4573184 -Total Bytes 878040 -BPFP 1.5360 bits/point -EBPFP 3.0720 equivalent bits/point -MSE 0.625604 ----------------------- -------------------------------------------------------- -Time: 0.979s Load: 0.018s, Pack+Encode: 0.365s, Decode+Unpack: 0.595s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 319, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 319, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6256 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample81-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample81-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample82-layer4-item1.zst (92/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample82-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 353, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.020s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 353, 128) -Output shape: (1, 353, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) -> torch.Size([1, 1, 353, 1024]) - layer.4.output: torch.Size([1, 353, 4096]) -> torch.Size([1, 1, 353, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 33,104B, BPFP=0.7326 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 83,376B, BPFP=1.8453 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,776B, BPFP=1.2787 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 88,688B, BPFP=1.9628 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,548B, BPFP=1.4507 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 90,556B, BPFP=2.0042 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,604B, BPFP=1.4962 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 89,076B, BPFP=1.9714 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,040B, BPFP=1.3067 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 91,308B, BPFP=2.0208 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 264,716B, BPFP=1.4647 -⌛️ [2/4] FRONTEND: Frontend time: 0.416s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.output: torch.Size([1, 353, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.696s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 353, 128]) - layer.0.v_cache: torch.Size([1, 8, 353, 128]) - layer.1.k_cache: torch.Size([1, 8, 353, 128]) - layer.1.v_cache: torch.Size([1, 8, 353, 128]) - layer.2.k_cache: torch.Size([1, 8, 353, 128]) - layer.2.v_cache: torch.Size([1, 8, 353, 128]) - layer.3.k_cache: torch.Size([1, 8, 353, 128]) - layer.3.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.k_cache: torch.Size([1, 8, 353, 128]) - layer.4.v_cache: torch.Size([1, 8, 353, 128]) - layer.4.output: torch.Size([1, 353, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02381471 6.54877210 - layer.0.v_cache 0.00000027 0.00023691 - layer.1.k_cache 0.00292436 0.62275713 - layer.1.v_cache 0.00000085 0.00086310 - layer.2.k_cache 0.00118067 0.38995292 - layer.2.v_cache 0.00000120 0.00124033 - layer.3.k_cache 0.00130453 0.44241225 - layer.3.v_cache 0.00000223 0.00200743 - layer.4.k_cache 0.00350570 0.75338598 - layer.4.v_cache 0.00000328 0.00341267 - layer.4.output 0.00016368 0.05874627 - ------------------------------------------------------------------------------------- - TOTAL 0.00238518 0.64285899 - (elements=5,060,608) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5060608 -Total Bytes 990792 -BPFP 1.5663 bits/point -EBPFP 3.1326 equivalent bits/point -MSE 0.642859 ----------------------- -------------------------------------------------------- -Time: 1.132s Load: 0.020s, Pack+Encode: 0.416s, Decode+Unpack: 0.696s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 353, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 353, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6429 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample82-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample82-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample83-layer4-item1.zst (93/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample83-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.020s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 344, 128) -Output shape: (1, 344, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) -> torch.Size([1, 1, 344, 1024]) - layer.4.output: torch.Size([1, 344, 4096]) -> torch.Size([1, 1, 344, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 31,516B, BPFP=0.7158 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,920B, BPFP=1.8605 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,496B, BPFP=1.3058 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,952B, BPFP=1.9975 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,172B, BPFP=1.4801 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,596B, BPFP=2.0348 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,232B, BPFP=1.5269 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,536B, BPFP=2.0107 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 58,448B, BPFP=1.3274 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,276B, BPFP=2.0502 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 260,576B, BPFP=1.4795 -⌛️ [2/4] FRONTEND: Frontend time: 0.426s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.700s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 344, 128]) - layer.0.v_cache: torch.Size([1, 8, 344, 128]) - layer.1.k_cache: torch.Size([1, 8, 344, 128]) - layer.1.v_cache: torch.Size([1, 8, 344, 128]) - layer.2.k_cache: torch.Size([1, 8, 344, 128]) - layer.2.v_cache: torch.Size([1, 8, 344, 128]) - layer.3.k_cache: torch.Size([1, 8, 344, 128]) - layer.3.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.k_cache: torch.Size([1, 8, 344, 128]) - layer.4.v_cache: torch.Size([1, 8, 344, 128]) - layer.4.output: torch.Size([1, 344, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02504009 7.04530938 - layer.0.v_cache 0.00000027 0.00024188 - layer.1.k_cache 0.00288154 0.61535192 - layer.1.v_cache 0.00000082 0.00088066 - layer.2.k_cache 0.00115445 0.39103867 - layer.2.v_cache 0.00000117 0.00124884 - layer.3.k_cache 0.00131257 0.43099936 - layer.3.v_cache 0.00000224 0.00205381 - layer.4.k_cache 0.00356118 0.74687940 - layer.4.v_cache 0.00000339 0.00345493 - layer.4.output 0.00014496 0.05694930 - ------------------------------------------------------------------------------------- - TOTAL 0.00246697 0.67608972 - (elements=4,931,584) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4931584 -Total Bytes 978720 -BPFP 1.5877 bits/point -EBPFP 3.1754 equivalent bits/point -MSE 0.676090 ----------------------- -------------------------------------------------------- -Time: 1.145s Load: 0.020s, Pack+Encode: 0.426s, Decode+Unpack: 0.700s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 344, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 344, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6761 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample83-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample83-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample85-layer4-item1.zst (94/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample85-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 321, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 321, 128) -Output shape: (1, 321, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.0.v_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.1.k_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.1.v_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.2.k_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.2.v_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.3.k_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.3.v_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.4.k_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.4.v_cache: torch.Size([1, 8, 321, 128]) -> torch.Size([1, 1, 321, 1024]) - layer.4.output: torch.Size([1, 321, 4096]) -> torch.Size([1, 1, 321, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,352B, BPFP=0.7144 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 76,612B, BPFP=1.8646 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 53,240B, BPFP=1.2958 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 82,456B, BPFP=2.0068 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 60,520B, BPFP=1.4729 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 84,248B, BPFP=2.0504 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 62,620B, BPFP=1.5240 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 82,828B, BPFP=2.0159 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 54,812B, BPFP=1.3340 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 84,704B, BPFP=2.0615 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 237,444B, BPFP=1.4447 -⌛️ [2/4] FRONTEND: Frontend time: 0.414s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 321, 128]) - layer.0.v_cache: torch.Size([1, 8, 321, 128]) - layer.1.k_cache: torch.Size([1, 8, 321, 128]) - layer.1.v_cache: torch.Size([1, 8, 321, 128]) - layer.2.k_cache: torch.Size([1, 8, 321, 128]) - layer.2.v_cache: torch.Size([1, 8, 321, 128]) - layer.3.k_cache: torch.Size([1, 8, 321, 128]) - layer.3.v_cache: torch.Size([1, 8, 321, 128]) - layer.4.k_cache: torch.Size([1, 8, 321, 128]) - layer.4.v_cache: torch.Size([1, 8, 321, 128]) - layer.4.output: torch.Size([1, 321, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.703s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 321, 128]) - layer.0.v_cache: torch.Size([1, 8, 321, 128]) - layer.1.k_cache: torch.Size([1, 8, 321, 128]) - layer.1.v_cache: torch.Size([1, 8, 321, 128]) - layer.2.k_cache: torch.Size([1, 8, 321, 128]) - layer.2.v_cache: torch.Size([1, 8, 321, 128]) - layer.3.k_cache: torch.Size([1, 8, 321, 128]) - layer.3.v_cache: torch.Size([1, 8, 321, 128]) - layer.4.k_cache: torch.Size([1, 8, 321, 128]) - layer.4.v_cache: torch.Size([1, 8, 321, 128]) - layer.4.output: torch.Size([1, 321, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02357271 6.66466563 - layer.0.v_cache 0.00000026 0.00022843 - layer.1.k_cache 0.00287206 0.60111073 - layer.1.v_cache 0.00000077 0.00081631 - layer.2.k_cache 0.00116438 0.39198638 - layer.2.v_cache 0.00000114 0.00119560 - layer.3.k_cache 0.00132116 0.42738689 - layer.3.v_cache 0.00000214 0.00192505 - layer.4.k_cache 0.00354096 0.76008539 - layer.4.v_cache 0.00000337 0.00337608 - layer.4.output 0.00013764 0.05328407 - ------------------------------------------------------------------------------------- - TOTAL 0.00235925 0.64756520 - (elements=4,601,856) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4601856 -Total Bytes 908836 -BPFP 1.5799 bits/point -EBPFP 3.1599 equivalent bits/point -MSE 0.647565 ----------------------- -------------------------------------------------------- -Time: 1.135s Load: 0.018s, Pack+Encode: 0.414s, Decode+Unpack: 0.703s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 321, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 321, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6476 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample85-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample85-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample86-layer4-item1.zst (95/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample86-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 329, 128) -Output shape: (1, 329, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.output: torch.Size([1, 329, 4096]) -> torch.Size([1, 1, 329, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,096B, BPFP=0.7147 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,752B, BPFP=1.8701 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,992B, BPFP=1.3059 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 84,796B, BPFP=2.0136 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,444B, BPFP=1.4828 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,620B, BPFP=2.0569 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,292B, BPFP=1.5267 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,296B, BPFP=2.0255 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,140B, BPFP=1.3331 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 87,204B, BPFP=2.0708 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 245,088B, BPFP=1.4550 -⌛️ [2/4] FRONTEND: Frontend time: 0.439s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.711s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02395787 7.18102248 - layer.0.v_cache 0.00000026 0.00022997 - layer.1.k_cache 0.00285945 0.64108462 - layer.1.v_cache 0.00000078 0.00086278 - layer.2.k_cache 0.00114467 0.39605110 - layer.2.v_cache 0.00000113 0.00122808 - layer.3.k_cache 0.00130198 0.43502650 - layer.3.v_cache 0.00000219 0.00199924 - layer.4.k_cache 0.00350851 0.78210023 - layer.4.v_cache 0.00000338 0.00347862 - layer.4.output 0.00014624 0.05633819 - ------------------------------------------------------------------------------------- - TOTAL 0.00238323 0.69060260 - (elements=4,716,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4716544 -Total Bytes 935720 -BPFP 1.5871 bits/point -EBPFP 3.1743 equivalent bits/point -MSE 0.690603 ----------------------- -------------------------------------------------------- -Time: 1.167s Load: 0.018s, Pack+Encode: 0.439s, Decode+Unpack: 0.711s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6906 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample86-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample86-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample87-layer4-item1.zst (96/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample87-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 350, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.018s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 350, 128) -Output shape: (1, 350, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.0.v_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.1.k_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.1.v_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.2.k_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.2.v_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.3.k_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.3.v_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.4.k_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.4.v_cache: torch.Size([1, 8, 350, 128]) -> torch.Size([1, 1, 350, 1024]) - layer.4.output: torch.Size([1, 350, 4096]) -> torch.Size([1, 1, 350, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,272B, BPFP=0.6757 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 81,572B, BPFP=1.8208 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 57,724B, BPFP=1.2885 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 87,352B, BPFP=1.9498 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 65,688B, BPFP=1.4663 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 89,280B, BPFP=1.9929 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 67,868B, BPFP=1.5149 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 88,312B, BPFP=1.9712 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 59,288B, BPFP=1.3234 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 90,512B, BPFP=2.0204 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 253,352B, BPFP=1.4138 -⌛️ [2/4] FRONTEND: Frontend time: 0.428s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 350, 128]) - layer.0.v_cache: torch.Size([1, 8, 350, 128]) - layer.1.k_cache: torch.Size([1, 8, 350, 128]) - layer.1.v_cache: torch.Size([1, 8, 350, 128]) - layer.2.k_cache: torch.Size([1, 8, 350, 128]) - layer.2.v_cache: torch.Size([1, 8, 350, 128]) - layer.3.k_cache: torch.Size([1, 8, 350, 128]) - layer.3.v_cache: torch.Size([1, 8, 350, 128]) - layer.4.k_cache: torch.Size([1, 8, 350, 128]) - layer.4.v_cache: torch.Size([1, 8, 350, 128]) - layer.4.output: torch.Size([1, 350, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.709s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 350, 128]) - layer.0.v_cache: torch.Size([1, 8, 350, 128]) - layer.1.k_cache: torch.Size([1, 8, 350, 128]) - layer.1.v_cache: torch.Size([1, 8, 350, 128]) - layer.2.k_cache: torch.Size([1, 8, 350, 128]) - layer.2.v_cache: torch.Size([1, 8, 350, 128]) - layer.3.k_cache: torch.Size([1, 8, 350, 128]) - layer.3.v_cache: torch.Size([1, 8, 350, 128]) - layer.4.k_cache: torch.Size([1, 8, 350, 128]) - layer.4.v_cache: torch.Size([1, 8, 350, 128]) - layer.4.output: torch.Size([1, 350, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02252116 6.42949009 - layer.0.v_cache 0.00000028 0.00022995 - layer.1.k_cache 0.00283737 0.62096675 - layer.1.v_cache 0.00000078 0.00077756 - layer.2.k_cache 0.00114986 0.38491342 - layer.2.v_cache 0.00000111 0.00112338 - layer.3.k_cache 0.00128401 0.43009129 - layer.3.v_cache 0.00000216 0.00184286 - layer.4.k_cache 0.00352053 0.74453055 - layer.4.v_cache 0.00000313 0.00319126 - layer.4.output 0.00014459 0.05491803 - ------------------------------------------------------------------------------------- - TOTAL 0.00227848 0.63120209 - (elements=5,017,600) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 5017600 -Total Bytes 971220 -BPFP 1.5485 bits/point -EBPFP 3.0970 equivalent bits/point -MSE 0.631202 ----------------------- -------------------------------------------------------- -Time: 1.156s Load: 0.018s, Pack+Encode: 0.428s, Decode+Unpack: 0.709s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 350, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 350, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6312 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample87-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample87-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample89-layer4-item1.zst (97/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample89-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 323, 128) -Output shape: (1, 323, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) -> torch.Size([1, 1, 323, 1024]) - layer.4.output: torch.Size([1, 323, 4096]) -> torch.Size([1, 1, 323, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,976B, BPFP=0.7250 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 77,096B, BPFP=1.8647 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,164B, BPFP=1.3101 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 83,204B, BPFP=2.0125 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 61,288B, BPFP=1.4824 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 85,000B, BPFP=2.0559 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 63,400B, BPFP=1.5335 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 83,800B, BPFP=2.0269 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 55,016B, BPFP=1.3307 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 85,372B, BPFP=2.0649 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 240,892B, BPFP=1.4566 -⌛️ [2/4] FRONTEND: Frontend time: 0.436s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.703s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 323, 128]) - layer.0.v_cache: torch.Size([1, 8, 323, 128]) - layer.1.k_cache: torch.Size([1, 8, 323, 128]) - layer.1.v_cache: torch.Size([1, 8, 323, 128]) - layer.2.k_cache: torch.Size([1, 8, 323, 128]) - layer.2.v_cache: torch.Size([1, 8, 323, 128]) - layer.3.k_cache: torch.Size([1, 8, 323, 128]) - layer.3.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.k_cache: torch.Size([1, 8, 323, 128]) - layer.4.v_cache: torch.Size([1, 8, 323, 128]) - layer.4.output: torch.Size([1, 323, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02344667 7.23781187 - layer.0.v_cache 0.00000026 0.00023034 - layer.1.k_cache 0.00293942 0.63698963 - layer.1.v_cache 0.00000088 0.00086786 - layer.2.k_cache 0.00118806 0.40355132 - layer.2.v_cache 0.00000115 0.00122024 - layer.3.k_cache 0.00133183 0.44056555 - layer.3.v_cache 0.00000219 0.00199687 - layer.4.k_cache 0.00365354 0.77477186 - layer.4.v_cache 0.00000330 0.00337169 - layer.4.output 0.00014070 0.06215399 - ------------------------------------------------------------------------------------- - TOTAL 0.00236644 0.69642808 - (elements=4,630,528) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4630528 -Total Bytes 919208 -BPFP 1.5881 bits/point -EBPFP 3.1762 equivalent bits/point -MSE 0.696428 ----------------------- -------------------------------------------------------- -Time: 1.157s Load: 0.017s, Pack+Encode: 0.436s, Decode+Unpack: 0.703s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 323, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 323, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6964 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample89-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample89-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample9-layer4-item1.zst (98/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample9-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 419, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.022s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 419, 128) -Output shape: (1, 419, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.0.v_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.1.k_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.1.v_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.2.k_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.2.v_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.3.k_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.3.v_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.4.k_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.4.v_cache: torch.Size([1, 8, 419, 128]) -> torch.Size([1, 1, 419, 1024]) - layer.4.output: torch.Size([1, 419, 4096]) -> torch.Size([1, 1, 419, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 36,192B, BPFP=0.6748 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 96,252B, BPFP=1.7947 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 68,432B, BPFP=1.2760 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 103,340B, BPFP=1.9268 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 77,428B, BPFP=1.4437 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 105,216B, BPFP=1.9618 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 80,024B, BPFP=1.4921 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 104,248B, BPFP=1.9438 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 70,280B, BPFP=1.3104 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 107,036B, BPFP=1.9957 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 296,608B, BPFP=1.3826 -⌛️ [2/4] FRONTEND: Frontend time: 0.489s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 419, 128]) - layer.0.v_cache: torch.Size([1, 8, 419, 128]) - layer.1.k_cache: torch.Size([1, 8, 419, 128]) - layer.1.v_cache: torch.Size([1, 8, 419, 128]) - layer.2.k_cache: torch.Size([1, 8, 419, 128]) - layer.2.v_cache: torch.Size([1, 8, 419, 128]) - layer.3.k_cache: torch.Size([1, 8, 419, 128]) - layer.3.v_cache: torch.Size([1, 8, 419, 128]) - layer.4.k_cache: torch.Size([1, 8, 419, 128]) - layer.4.v_cache: torch.Size([1, 8, 419, 128]) - layer.4.output: torch.Size([1, 419, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.789s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 419, 128]) - layer.0.v_cache: torch.Size([1, 8, 419, 128]) - layer.1.k_cache: torch.Size([1, 8, 419, 128]) - layer.1.v_cache: torch.Size([1, 8, 419, 128]) - layer.2.k_cache: torch.Size([1, 8, 419, 128]) - layer.2.v_cache: torch.Size([1, 8, 419, 128]) - layer.3.k_cache: torch.Size([1, 8, 419, 128]) - layer.3.v_cache: torch.Size([1, 8, 419, 128]) - layer.4.k_cache: torch.Size([1, 8, 419, 128]) - layer.4.v_cache: torch.Size([1, 8, 419, 128]) - layer.4.output: torch.Size([1, 419, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02248870 6.81283038 - layer.0.v_cache 0.00000028 0.00022886 - layer.1.k_cache 0.00282601 0.60933520 - layer.1.v_cache 0.00000075 0.00076714 - layer.2.k_cache 0.00117249 0.37787721 - layer.2.v_cache 0.00000110 0.00109113 - layer.3.k_cache 0.00129145 0.41793612 - layer.3.v_cache 0.00000206 0.00181983 - layer.4.k_cache 0.00360397 0.73873020 - layer.4.v_cache 0.00000307 0.00313917 - layer.4.output 0.00013808 0.05160209 - ------------------------------------------------------------------------------------- - TOTAL 0.00228159 0.65501169 - (elements=6,006,784) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 6006784 -Total Bytes 1145056 -BPFP 1.5250 bits/point -EBPFP 3.0500 equivalent bits/point -MSE 0.655012 ----------------------- -------------------------------------------------------- -Time: 1.300s Load: 0.022s, Pack+Encode: 0.489s, Decode+Unpack: 0.789s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 419, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 419, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6550 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample9-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample9-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample93-layer4-item1.zst (99/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample93-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 336, 128) -Output shape: (1, 336, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) -> torch.Size([1, 1, 336, 1024]) - layer.4.output: torch.Size([1, 336, 4096]) -> torch.Size([1, 1, 336, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 30,828B, BPFP=0.7168 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 80,008B, BPFP=1.8603 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 56,100B, BPFP=1.3044 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 86,160B, BPFP=2.0033 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 63,612B, BPFP=1.4791 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 87,852B, BPFP=2.0427 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 65,584B, BPFP=1.5249 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 86,468B, BPFP=2.0105 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 57,352B, BPFP=1.3335 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 88,616B, BPFP=2.0605 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 250,904B, BPFP=1.4585 -⌛️ [2/4] FRONTEND: Frontend time: 0.428s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.708s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 336, 128]) - layer.0.v_cache: torch.Size([1, 8, 336, 128]) - layer.1.k_cache: torch.Size([1, 8, 336, 128]) - layer.1.v_cache: torch.Size([1, 8, 336, 128]) - layer.2.k_cache: torch.Size([1, 8, 336, 128]) - layer.2.v_cache: torch.Size([1, 8, 336, 128]) - layer.3.k_cache: torch.Size([1, 8, 336, 128]) - layer.3.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.k_cache: torch.Size([1, 8, 336, 128]) - layer.4.v_cache: torch.Size([1, 8, 336, 128]) - layer.4.output: torch.Size([1, 336, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02369028 7.24937148 - layer.0.v_cache 0.00000026 0.00023443 - layer.1.k_cache 0.00291391 0.63770580 - layer.1.v_cache 0.00000078 0.00087453 - layer.2.k_cache 0.00115383 0.39223112 - layer.2.v_cache 0.00000117 0.00123653 - layer.3.k_cache 0.00133386 0.43636799 - layer.3.v_cache 0.00000214 0.00197328 - layer.4.k_cache 0.00353874 0.76934397 - layer.4.v_cache 0.00000342 0.00345982 - layer.4.output 0.00014590 0.05712607 - ------------------------------------------------------------------------------------- - TOTAL 0.00237300 0.69437880 - (elements=4,816,896) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4816896 -Total Bytes 953484 -BPFP 1.5836 bits/point -EBPFP 3.1671 equivalent bits/point -MSE 0.694379 ----------------------- -------------------------------------------------------- -Time: 1.153s Load: 0.017s, Pack+Encode: 0.428s, Decode+Unpack: 0.708s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 336, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 336, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6944 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample93-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample93-layer4-item1.zst - - 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample97-layer4-item1.zst (100/100) - -[1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample97-layer4-item1.zst... - -Original data structure: -root: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -⌛️ [1/4] FRONTEND: Load time: 0.017s - ------------------------------------------------------------- -Qwen Features Summary ------------------------------------------------------------- -Number of layers: 5 -Layer indices: [0, 1, 2, 3, 4] -Last layer index: 4 -Cache shape: (1, 8, 329, 128) -Output shape: (1, 329, 4096) -Data type: torch.bfloat16 ------------------------------------------------------------- - -[2/4] FRONTEND: Pack + Encode (strategy: individual)... - IndividualPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) -> torch.Size([1, 1, 329, 1024]) - layer.4.output: torch.Size([1, 329, 4096]) -> torch.Size([1, 1, 329, 4096]) - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - layer.0.k_cache: 29,752B, BPFP=0.7065 - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - layer.0.v_cache: 78,804B, BPFP=1.8713 - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - layer.1.k_cache: 54,972B, BPFP=1.3054 - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - layer.1.v_cache: 85,064B, BPFP=2.0199 - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - layer.2.k_cache: 62,696B, BPFP=1.4888 - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - layer.2.v_cache: 86,336B, BPFP=2.0502 - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - layer.3.k_cache: 64,380B, BPFP=1.5288 - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - layer.3.v_cache: 85,036B, BPFP=2.0193 - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - layer.4.k_cache: 56,048B, BPFP=1.3309 - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - layer.4.v_cache: 86,968B, BPFP=2.0652 - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - layer.4.output: 242,704B, BPFP=1.4408 -⌛️ [2/4] FRONTEND: Frontend time: 0.423s (Pack+Encode) - -[3/4] BACKEND: Decode + Unpack... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) -⌛️ [3/4] BACKEND: Backend time: 0.704s - -[4/4] METRICS: Computing MSE Breakdown... - Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache - Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache - Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache - Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache - Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache - Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache - Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache - Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache - Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache - Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache - Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output - IndividualUnPacker: - layer.0.k_cache: torch.Size([1, 8, 329, 128]) - layer.0.v_cache: torch.Size([1, 8, 329, 128]) - layer.1.k_cache: torch.Size([1, 8, 329, 128]) - layer.1.v_cache: torch.Size([1, 8, 329, 128]) - layer.2.k_cache: torch.Size([1, 8, 329, 128]) - layer.2.v_cache: torch.Size([1, 8, 329, 128]) - layer.3.k_cache: torch.Size([1, 8, 329, 128]) - layer.3.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.k_cache: torch.Size([1, 8, 329, 128]) - layer.4.v_cache: torch.Size([1, 8, 329, 128]) - layer.4.output: torch.Size([1, 329, 4096]) - Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): - Key Quant-MSE Total-MSE - ------------------------------------------------------------------------------------- - layer.0.k_cache 0.02541956 7.23644166 - layer.0.v_cache 0.00000027 0.00023273 - layer.1.k_cache 0.00292153 0.62546713 - layer.1.v_cache 0.00000078 0.00086326 - layer.2.k_cache 0.00117184 0.39904386 - layer.2.v_cache 0.00000114 0.00122204 - layer.3.k_cache 0.00131754 0.43041487 - layer.3.v_cache 0.00000217 0.00200639 - layer.4.k_cache 0.00353394 0.78216247 - layer.4.v_cache 0.00000342 0.00342203 - layer.4.output 0.00014204 0.05483840 - ------------------------------------------------------------------------------------- - TOTAL 0.00249574 0.69290214 - (elements=4,716,544) ----------------------- -------------------------------------------------------- -SAMPLE-WISE STATISTICS ----------------------- -------------------------------------------------------- -Handler qwen -Strategy individual -Architecture hyperprior-featurecoding ----------------------- -------------------------------------------------------- -Total Elements 4716544 -Total Bytes 932760 -BPFP 1.5821 bits/point -EBPFP 3.1642 equivalent bits/point -MSE 0.692902 ----------------------- -------------------------------------------------------- -Time: 1.144s Load: 0.017s, Pack+Encode: 0.423s, Decode+Unpack: 0.704s ----------------------- -------------------------------------------------------- -Restored Feature Format: [dict] with 3 keys - key['output']: [Tensor] shape=torch.Size([1, 329, 4096]), dtype=torch.bfloat16, device=cpu - key['key']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - key['value']: [list] with 5 items - item[0]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[1]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[2]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[3]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu - item[4]: [Tensor] shape=torch.Size([1, 8, 329, 128]), dtype=torch.bfloat16, device=cpu -💾 Converting with 0.6929 MSE: - from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_hellaswag/sample97-layer4-item1.zst - to output-fixed/qwen/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample97-layer4-item1.zst ------------------------- ---------------------------- -TOTAL PROCESSING SUMMARY ------------------------- ---------------------------- -Total files 100 -Avg BPFP 1.5701 bits/point -Avg EBPFP 3.1402 equivalent bits/point -Avg MSE 0.669145 -Avg Time 1.138s ------------------------- ---------------------------- +version https://git-lfs.github.com/spec/v1 +oid sha256:33ddc9ea37fec193b195b8ad1dc8fecc1b2e32e9f9c8b22d435d29ff230ec62b +size 1124962 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample0-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample0-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..567781422366bc295cd42583d3c35daf6641f2f4 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample0-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0a302fbcbe95f9d016cd680332a373545bac98aaf927df29624186752604d254 +size 6502453 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample1-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample1-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..cacf8fa1361944a3540369a90843434f12672a02 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample1-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07f36037aca29ca0005eb377fa99592f5d35d09fee58c566b207d56f61683e1b +size 6062215 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample10-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample10-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b34241f2115ec8cf1f23dcda055e3fa410e72959 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample10-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d344ff2539885058fd0a6884a72ab657320573a4545cafd1beaa4fc85284f29e +size 6638245 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample100-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample100-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..54e6a273488d96e6553829acfdd70460e55a4486 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample100-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4152e66018a6416f3057a21600a0ce20c9e8d7fbf1769288f0c9ffc9c928c3e9 +size 5554792 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample102-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample102-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f773e1082a037150ac746f76caedcc435547f353 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample102-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21dd893043f4a485e15ec39cfe1a4e0c0169456e0a609ec2e2ed01e8def1b7a7 +size 6346358 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample105-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample105-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c90fa90d85b576cea5dda4af36741c7f0ca14447 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample105-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41fa4cd60d019edd3b31907fe0c04fe3db1a7a4cbd17ab44ad64aef11942cd5c +size 5894430 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample108-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample108-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2485c0cc93a2c67617c045497b0ffdd01461d599 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample108-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7e99501b9014017971f1f577a488f927e8c394b26f001298e3d1c18908ece78 +size 5581787 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample11-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample11-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..652356f34fe84727120aebe887f401e26e975ad0 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample11-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce05b4ad23cfb430609c1d7857622e11eea0a988a3bb724d19315f18b579aaa2 +size 6187711 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample112-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample112-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f81af0fdeb9405875dd7aa40948cbe713f34d7de --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample112-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96b7d6dcfbab7d8c4bce72a5de7333eb812b44470fe35919f302794ff3a6bc8f +size 5601505 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample114-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample114-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1e76a1074334e38e81f21e3bb5fe39a3881dfd55 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample114-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f50fce92525cfba8c1f435c2e291fe4b53be96bfe7284993dcf9af3f56b0e5ac +size 5815297 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample119-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample119-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..dea9bef0b2404efee498821f877e6673469ff243 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample119-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b6114e35e79b1ea82fd3ff147813d9eb26279059632d5780f6c34866109f610 +size 6297489 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample12-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample12-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3459415bd5b9d756a31d33ba848b01d125989db9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample12-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6c463e940217d577cda4b63d94e32d3462c41ed89510c86fb134b8ef190d6319 +size 6295895 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample13-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample13-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5011c2228268247892889e8d98716930c8966c3f --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample13-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:058b3cf7ef4b44c56065859e367eff66e8616241c1cc7330db060f07cdd3c7ea +size 5918132 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample132-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample132-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7146d6d642cd5fdce416f0dc32edec914ff8a41d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample132-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a031e1a54facc2e7c2377ec3a857f570080ea816e3763fdff1391ffe00e17cc +size 6098905 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample135-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample135-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7d3e693f0f5b6bf5ce52edad016445a7779fe6d2 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample135-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70c456ff4e7fdc77bfa1bb914857620e9cf4d26887219eade186aaf480629808 +size 6010648 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample14-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample14-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..54a893d6156e833318667f55b4b28019d010ea25 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample14-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce88309057692745b2643efa583f0e3afb4e8f38a9c5de0fb815f5404bd73d43 +size 6310571 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample15-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample15-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..fd41242b20aa7f8bd7d2f18d70322ff4e89dcf7f --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample15-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24e0fcfd2b48445e64bb4e18c661217653ad330b3bfd2ebc0ca64a244f768a3b +size 6318691 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample16-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample16-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0c7243e030abd6b324dc4e2afaeca4981d125758 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample16-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a5766fea70f3a47300b33a4a4981c7add1cc88b20f2173a9b55ac7639d43c7a +size 6095875 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample165-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample165-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2119f6f3e1f0147e12cbfec626aaa3d4bb9853e0 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample165-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c68754157701ce3c0262edb9f68ee92d6067edd3d0a69afcd1b13e285850dbb7 +size 5905029 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample17-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample17-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..68d81015d3f7191100f449aef2cf22ca87058329 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample17-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31357e9f02f754a7723c58798ea5347e0cc1c588a1c04bbc36b8a52829ea3b00 +size 6151857 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample18-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample18-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7c3480f2b43c578b209f06944a65325bba3ab128 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample18-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13c5587903f45d17d9a9f828907f48e8bb118acb603a8a6a746a3865fe62b3a6 +size 6086815 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample19-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample19-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f027a33ea3f6f57775cc8003069a9159f00f6fa6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample19-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94a1d599299ca975139ae9b5cfaad4ff48f596e480a2957cd1387cf6c002d014 +size 5991652 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample2-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample2-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3e14ef88993dcb16bb5f17ca1e9982c98dcaf441 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample2-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b812fa009c3a00fc7c0e04fb591c0d7f5cfe84833ba31ca9289c348ff4ddbaae +size 6058305 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample20-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample20-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0d6278aee49f3a68cd5130f012ff84475973d7c6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample20-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1339dcbdcfa1bd3a3a43110c15350bc737fec59cb5cb307b2f084998aa592264 +size 5919128 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample21-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample21-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b16cede2f76de72957feacc04667c3b2ab8f81ec --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample21-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2fd6398ff1456c22f90a56fc62f05cb9b09be21fdea8778949474bf97fe8239 +size 6915825 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample22-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample22-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..818c7f41d4fbdbc070bd428e6c0288fe3f4d4bef --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample22-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a69516b1d4600d54c82b77f12d8b877583df0fb46ee36c4ef1b4edbb8cfd088 +size 6105081 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample23-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample23-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..cd767db9aee2792a5dfd5f6d86123582dcf18f6d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample23-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f75215eae53cf10bc11f7db7ba35168673e7b9c96a930c3704350afef793d648 +size 5563728 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample24-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample24-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..06501d3db8934ac57c9da200b54ea13184f67188 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample24-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10ea6f22543ddb24bb9e08cd2aadd810513d889806dd8e0124090bedbb383f31 +size 6113595 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample25-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample25-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f7c411a00d19203d15d41b30917d6ec6acd61acf --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample25-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:936225ac0d21b971dcdff6ec3c4cd80177291ba50a39e74ddb32a9f480de4c3a +size 6067863 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample26-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample26-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b215edcce86f0289faba654027ec6278382288f7 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample26-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48fa3b4bc41749eebed41c83393a1bf91ef0c5eb4a61e4c7eaba9ed056bd08bf +size 5814408 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample27-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample27-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5633e282871868bb16ee9960623c190d0823555e --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample27-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9446c9cb6445b1907f4bedf8b69e8e7efef0d16f50000b8ede4d1d856d34ef7 +size 5876668 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample28-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample28-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..14facb98e3bc31b368a609aeb53b2772705ccaf7 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample28-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ece432ca81396341eaf03c1c008c0a0e8ea9d49cdb1bddf7401de4e3b28e14d +size 5815668 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample29-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample29-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..678c57e85f5acf9d637e59be95506e3adacf02d9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample29-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:270f0c32c946ee1805ed90459dc112bda742c696beab0da3126b742883493734 +size 5798109 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample3-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample3-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..79e71ff640051c3f84c149e2f9c247dd2d77544b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample3-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc5e3b7053f452f3dba6ca17743a6650e5af68583830a932f1f9e0d4d8436b63 +size 6428977 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample30-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample30-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..24088188d718ceee5577eb32fe7e66e8df86ea0a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample30-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94cf3ef6b1b18fae568516a5097c40f6d1073731761f52a778eabd46acadb783 +size 5903018 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample31-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample31-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e4db0007e7b4cc95ae2d0bac4b08f2ceaf750355 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample31-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69852a93eebd4e1478cc7787eff41fc3c1b430804b3b0082c443facda047cd7b +size 5890460 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample32-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample32-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..32828ff35ae2862a267ac34c2d37c6a27d02ed9a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample32-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43c392981bcd5de45a40bb05ca39b7113efe394d303b24af7ab1e75c52a262a1 +size 5973098 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample33-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample33-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..45a8479b615c91af2e27a141d5b5bc9c883e1434 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample33-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b02f726fb31e3795ac6697a56d1b2e8f3a8eb71bbebc0fce0b0cc8e771d08a4f +size 6357015 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample34-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample34-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..e183d930437a59e3561a6e4eaafeb088c0013c24 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample34-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:93a5228e9a2d2aa76a575b814da53ffa634d05331ec8635953191683d98e2d09 +size 5614484 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample35-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample35-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0aeeea1cee003c7d91c151702c05f86d07730b25 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample35-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89f5a52116e7a05b52df94d067685cd343fe8d5cdd407c74628578dd1ac79f1a +size 6016840 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample36-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample36-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f03ab154bcd79be058f95de781a1c28cf6b06228 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample36-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78610b83fabbaa7d1525407af2a22a8010e300346eecf7bd58dab93d03824069 +size 6045565 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample37-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample37-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3a16d85da6245c3ceefbe8303919c08521806ebf --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample37-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c506dba1e43770fd98376d8e2028773a9b17268328a1409da4fc0684a591d5c5 +size 5850384 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample38-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample38-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d26d199e243c7ef1fda64dd9434de1087490c830 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample38-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6d51c63799c0a367b50b384f853bd43f4d643f98fb1c4129a47ac0934a734e1b +size 6258884 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample39-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample39-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..25eee7e1407426e0abe7534c4f063d9a398295f7 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample39-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0db3038103e0a89647877682c253276c89bebc924848150f4fde5e2d08839bfc +size 5822225 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample4-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample4-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..a3cadd971b18a3567048728d575884e1a81f1655 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample4-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15efa7dfe6d650f179e9fc59d4ed4b1846aeaad8d03720aa66ed4a6cc623a718 +size 6079386 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample40-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample40-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..55e0616a9d7e87b27ace06e7189437cc4d8bfc47 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample40-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc9d8a9eab8ba5b0089b6b4210393de9acbe9e2c0d92d67e7b214cba1b40d932 +size 6054778 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample41-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample41-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..36666094ca31519cc538cc5d039b13ddc8fe6717 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample41-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1120a529e4ab523f8216247f6d1ee610e155a5800893d54a82c5ed1312540123 +size 5969587 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample42-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample42-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..10f55715a8bd6539506172db6195808025482a64 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample42-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f9b02aa7322bb8cd216cc70f8bfe970a285cd95c144b338d2f2078f466b5bed +size 5571869 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample43-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample43-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..14756c26aa894f6b608f1955d4e3dec702afef61 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample43-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05eb8d5927cff33c805e87b7d0b2604ae48b4d297a924d17f278224b681305df +size 6040220 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample44-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample44-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..60a3671677d9afc12081e01e6f489836686e3d51 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample44-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:260702ee6bfd14620bc84b90a42b1d45c8377661a5ab1e4234a85b9badda0147 +size 5722671 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample45-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample45-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d0d18c4743e046ec1163e12767bdba14e01a86b6 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample45-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51f9ea32b521d2066a881780d2c08c13a9d667e3003187c51553661e6971a6a7 +size 5901394 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample46-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample46-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f62d9115b3f5e533e7f702ed8d81f81f5e3e5690 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample46-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d71640ad777b1388858bf3c75910e70ae90f8d3299e3800d22b7d3c35025d2f6 +size 5697977 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample47-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample47-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..0430e437b3096a1fc58b927acae71ada2ad7c1eb --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample47-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b98fcd4e9049c2da0def9817166d6eaddddf03c85d5c3a893641f55442812ee +size 5849640 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample48-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample48-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4b4df40e81c532ace36f2a4d1684420cfb7fe6ba --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample48-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cbf5a8a5ea45b91d2fcc79b1e03f3f5448c9703d03ebddec4488999a661f105f +size 5624529 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample49-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample49-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..617bd3a788d4439ecc0f7c3b69a63e5a336067ed --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample49-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e9bf51fa3fb0f78df8a28d728726d7298b1e016436ca286fc6d0c4a0b984032 +size 5890352 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample5-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample5-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..ebd69f3996e530383b4bc4dd488ad5548b8b5fe8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample5-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3e6e0b3e26e83f9235e7fea27e7ac69f25fb5bbffb997f497b93a9e610a22a3 +size 6242262 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample50-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample50-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..6f050dd0e88e6223b331f84e91657110d1ea5bb4 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample50-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4005c9a2b81f19336fbc7ed2118af6986d4955c2492ff27f6dcd3f5e8933b76b +size 5765619 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample51-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample51-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..8ba49dd3e318f74b99697d844143f5ae2eb3694c --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample51-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e97a2b513a5fb403ff4e519a2f9fc1587f445a09ed49d5edb314d47e65ec4a25 +size 5522842 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample52-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample52-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..62979c08f1fc4ea96caae1f23ec5524bbe5dab8a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample52-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d46f221c275c779bd440ff013450ce3a86e5d37423203fba28498487e40f6a31 +size 5798462 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample53-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample53-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2262d94720f4657476c948b595bdcb228e458514 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample53-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ba9f27c84fa47346e0f8c6a519f4a8e7e4d95662f59aabd52928b52fe09ed1a +size 5810661 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample54-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample54-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..064174e4a6fd56d1d64e2d9a2c85782bf8a5908f --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample54-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e1d033f6a8f572cfe6a285493dbb7b0d18ddb9a60b42fba845c51a80eff4d81 +size 5650584 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample55-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample55-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5900ec17179c23d009ac108c003c55149235e383 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample55-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14f0f6abb37c8abb8591e34219f38453e637edb09f3ae7274c66ba2e5dfbb7d2 +size 5966599 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample56-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample56-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..6f2f48e779523ed5386beb8497eb11dccada205b --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample56-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d5daf2dc19309e1a0648b19d16fc70d70497cc7c5f79288103af2b759392121 +size 5531833 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample57-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample57-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d48736bea15e573b77f38f3e65da297294c6c2c7 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample57-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4120b905934a0ca2c75b8ce6e662a1f069238fa193875aaae866eebd2147c143 +size 6055986 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample58-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample58-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..227120b7e1ff0a0a060987b1e5e7d7fce93624f8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample58-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23f56b4309300e7a3084e2ce731c4eb25e5dd6b7fd10bab64c3e1c381a2ac743 +size 5443605 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample59-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample59-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..46a3018805c4b1d7f127dfa9e03a4cd191fa4d85 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample59-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e97ef27eac000fda01bf05fddefedd796bff50b0b1b6bd3dcf0c9fa116a566c +size 5962440 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample6-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample6-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c21fbf517b61fe1755d0a39449af4c571c256ac9 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample6-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c729171f4f9f20d714794471932462770af53b10ceecdb8a9d4b887eb6a13380 +size 6241522 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample60-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample60-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..766ea2d7ae9adda8775f189c0eaa7c4f81ffe084 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample60-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7ea87b8f31b195672629f2c1515914cd4399f914e4efbd6d9697552c1ac65d5 +size 5880884 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample61-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample61-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..12c68dd71a0d784fcbfee1fe9314c1a3ab1c8b49 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample61-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3702a464e081ab6021c085065315d69accb32c380eb6de2cefbfe9c2bb774293 +size 6002199 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample62-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample62-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..6ea9e3da1ac4e88916d8dda909b0b2a50b8d40ac --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample62-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8d6a6089439926f8627a6ea551852eb9f65c4e27248f183e97293134d125411 +size 5489423 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample63-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample63-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..9e956263be93dccb2cb25769c2c4e3908c2a8268 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample63-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69db23072077c02dc8404336547d89a740df045748edad0acb001b4332613e53 +size 6177076 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample64-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample64-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..da515ea713329122eb84c97f0f70ae410720fba2 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample64-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc4d641d31ccb3bfcf51fc71410ccb3c4f4e86dc70b58257f7f41e9597d31245 +size 5666044 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample65-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample65-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..181b004c5521389b7b80d8afac1b337892e0e634 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample65-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6afb17107ee31a856d7fdc0980fd08f9958113543533be7e7237d90cbb9b570d +size 6256014 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample66-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample66-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4e98ac79e97d14b47079ededa4dae25e9ad04142 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample66-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9d5b6b742a65524afd024b94e219ea7a452dad6f1b6b629582b67443e6bfae7 +size 6014475 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample67-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample67-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d0ab51d64f9b728d832cf9d4ff1b3cdf72d16f8d --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample67-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22eca5286efcdedfac774cda786e0f790e2101a04f9f4dd4da169c2cba9a718b +size 6343354 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample68-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample68-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..62b85f71b8fe4c97512078e1c3bae56d860f1afb --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample68-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79c97bea4e5828ac31592de25c0283ea74bf2ad80fe3315b0f1a17cb062b7655 +size 5185333 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample69-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample69-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..110e028fbbdf29b2e3bf28512755b6374fb09cf8 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample69-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:429356f3bdb6a886b4f33ec98ddf78a40fbb35d235f820cdfa1294dccdc002cb +size 5948466 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample7-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample7-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4df06cef8bb60cf4a3278d18ba07e8ddae3a4225 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample7-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39506985f9b0c6a1410c0fec241fcda174e114a9c4441726b68be6cf771d20ca +size 6001994 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample70-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample70-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..b8e7a86bf18490076d5d4e654107a3fd26993591 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample70-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdff78835c74a9fc169a8c9b34c6435ddfabc5c0b267ee62fa248f79725a96cf +size 5642649 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample71-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample71-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c0e6a5a102fb0c48d3cc555d459ef23b48b61b67 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample71-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ce54e1c3595684fadb0adace290d17b9a56fb5ea041ffc3827e388bc9d7fd6e7 +size 5772181 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample72-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample72-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..363cb3be114877e7f3de93caa94360b15d37bbd5 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample72-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6080e6c0c64863fbbb1d7f12cd78709bc30091d331cc6c962c59015217c4c5c9 +size 5765857 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample73-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample73-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..3b90692dd650e28dd3429985089a66192beab714 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample73-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:896e5d8216a5587836bf5f8343d24d2e775c9ac36482fe42dfe30cdc5b633562 +size 6189272 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample74-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample74-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..5185731bf8e419909ae8b3246f56d73d697c2ce1 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample74-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acf693d63896917e1e110fae678630420375f1959b23e41d9883c1a55b744d08 +size 5704251 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample75-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample75-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..33d7ac6e4a8ab90be5969991304ae859072cfcef --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample75-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a8e521dc90a74607f54d3d5c20715211fdd01a34fd051175f62cfd801a06b32 +size 5767048 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample76-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample76-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..40495a9bb50f61d73d285250fd661ce57d3cf828 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample76-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:17b5366f4872ab4ebaad8941dba295aa9d0e6e1bc4891565969e9399ada77f31 +size 5803147 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample77-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample77-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..8d02f86d0a9e56eedc8076ffc4a6ca77397de145 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample77-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f134fcb0d2cd08cc0292ca81f522464768a9b639d216f3fa73eb08cf98db9cd6 +size 5775558 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample78-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample78-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..82c92a20e78437f94ca511acd5cb8ed3606d9031 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample78-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3b418da6f131915b9b1b3249f8858780603782257fc41d8760f40a52626ae58 +size 5851129 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample79-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample79-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..81e3a6d4e71e0f590bd6df655c7a7fe3c354a808 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample79-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5424cc978cba9caf1eff7297ca98c143347bcda8adbee1099b68499ce1c26d3 +size 5352899 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample8-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample8-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..748858352415e8cd2f6c787eecabcfc0229117be --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample8-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfd0ce486df690b61b14f6608dec1ea0223662aefe3dc47a835ab7c8a27a3c93 +size 6179262 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample80-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample80-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2e2c4658d3532680b9ff8091956079d9d63ab9a1 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample80-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0ce89fb11e2ccf4168603a84521460de5858799ecb6dc1f0c9e8641f2742174 +size 5840654 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample81-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample81-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d7e3019c5a0e7837a581054a59582d575a50a840 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample81-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbcc34d2bce4a3fb423d80f8ebc5f28c2bbd36c3d410e5b38145dc20b18014af +size 5644054 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample82-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample82-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..35e24bcbe9dcf454e74d8f60c2a2e6e929181890 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample82-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04c8b233e4e796b5dedc2de92165075738839a3393d05b33bd5ec24db89c82f8 +size 6233606 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample83-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample83-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..d63f2026bad225003606e69feea458c5a113496a --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample83-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4527946170b56727e827725b400ac1f5d79aa2122608a5d86189a37d3cecdf13 +size 6069816 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample85-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample85-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..62ef7e0031f868fcca77a1638bf361deb418f0d4 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample85-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9765850b6b2bca7cc5dfac451d87e242604b180d516a14cf9b8bdbbcd2249ade +size 5649200 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample86-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample86-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..7b25b5a6930eb70cc5d1e7cfc850f43490ecd470 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample86-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb5ebe49b6e48d72708bb7d0c1044be65e5aef16f8d13873e9ba540e21ff8130 +size 5784863 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample87-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample87-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..f55cb95e081d509196ba9551a7902942d7acc4f5 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample87-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:836c662d09d4c09be9ca784055ce4d70a430cf58077f395775d23ba1c449a915 +size 6134026 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample89-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample89-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..4b62d48720015abfbf7ad0c35f42d4264995d894 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample89-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:473794eb15ca6e09c5569e668104db7ed78b4c110090e35a067c9714c8793dae +size 5679167 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample9-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample9-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..1f1b7045c7145073b4e14fb3946115646a9378a3 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample9-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6afe2347ef6b08aae479afcf1a1dad665552a316db19a00680854798140d011a +size 7319873 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample93-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample93-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..c135269442c56d46a691a1377a0a4e421907babd --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample93-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:88e228cac7d2ee7b549064eb7b75008199148c2849072c10d129c64e8c887c0c +size 5912109 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample97-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample97-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..390e78dd0a5feeac9ac079d0d29c92d12da8f859 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_hellaswag/sample97-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:66d685e7cd985a0d2d889c248260d4f3010fd26ff3ad29480577d19225b98486 +size 5787008 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample1-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample1-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..2a3d61b5b3535a211df248d7c9313ee487c5d570 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample1-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e1f31d7542a10efbd4d1e7009bb10ae3da15a6dcedcb223f8e91ee3a7831900 +size 4835944 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample10-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample10-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..bba4ec08562f05e165cf95e211b4e2f2b4ebbfec --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample10-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:96bfc425dda9046b809596c4e5b46268aaf0ca546c7d3ddd712fce32b4565108 +size 4971637 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample100-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample100-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..50e1e630ae35494a73c411218ded1937663eac91 --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample100-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fa54ab6bbaca95fdbc61ad38a530e5fa4687e0d81c5fb72f6ef918a095c47dd +size 2798906 diff --git a/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample102-layer4-item1.zst b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample102-layer4-item1.zst new file mode 100644 index 0000000000000000000000000000000000000000..a1b171d5c57fa00207f871482d2b2fa56168273e --- /dev/null +++ b/lambda0.01/hyperprior-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample102-layer4-item1.zst @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23d06ba308b11aebbd16cf1b149deeecbdc0138f155a657b8899d0a956b00183 +size 2672833