Publish qualified 45-second and four-minute quality bundle

README.md

@@ -14,73 +14,85 @@ See LICENSE and NOTICE. This conversion is not endorsed by NVIDIA.
 Choose one self-contained bundle:
 
 - `fast/`: full finite Fourier relative attention, 192 encoder positions,
-  fixed 15-second chunks, up to 128 independent chunks in the FP16 ANE decoder.
-- `quality/`: full sinusoidal relative attention, 3,776 encoder positions,
-  exact tiled convolution subsampling, fixed 300-second chunks and an FP32 CPU
-  decoder batching up to four independent chunks.
-
-Both use W8A16 encoder weights, retaining selected sensitive projections in
+  fixed 15-second chunks, and an FP16 ANE decoder batching up to 128 independent chunks.
+- `quality/`: full sinusoidal relative attention, shared weights for 576 and
+  3,008 encoder positions (45-second and four-minute inputs), exact tiled
+  convolution subsampling, fixed 240-second chunks, and an FP32 CPU decoder
+  batching up to four independent chunks. Select the smallest shape that fits
+  each chunk; short recordings use the 45-second shape.
+
+Both use W8A16 encoder weights with selected sensitive projections retained in
 FP16, and four concurrent encoder requests. Neither truncates attention within
-its chunk or drops input audio. `runtime.json` records the qualified scheduling
-policy. The two bundles trade speed for longer context; the labels describe the
-measured JFK result rather than a universal quality ranking.
+its chunk or drops input audio. `metadata.json` supplies available input shapes;
+`runtime.json` records the qualified scheduling policy. Weights are shared across
+static functions in each asset. The labels describe a measured speed/context
+tradeoff, not a universal quality ranking.
 
 Requires physical Apple silicon on macOS 27 or iOS 27 and a model-specific host
 runtime. Graphs accept model features and recurrent states, not audio files.
 The host must implement the frontend, greedy TDT loop and tokenizer described
-by metadata/sidecars. Weights and recurrent states are shared across static
-functions; recurrent histories remain independent between utterances.
-Vocabulary: 1024 nonblank tokens, blank ID
-1024, durations 0/1/2/3/4 and at most ten symbols per frame.
-For the larger V3 vocabulary the batched graph returns `(partition, local)` in
-two FP16 channels; reconstruct a token as `partition * 2048 + local`.
-Global IDs above 2,048 must not be transported as a single FP16 value.
-The scalar FP32 decoder returns a global token ID.
+by the metadata and sidecars. Vocabulary: 1,024 nonblank tokens, blank ID 1,024,
+durations 0/1/2/3/4, and at most ten symbols per frame. Recurrent state is independent
+between chunks. TDT emission frames and predicted durations support token/word
+timings at an 80 ms frame step; they are native alignments, not forced alignment.
 
 ## Measured performance
 
 M3 MacBook Air (16 GB), macOS 27 build 26A428. Medians after warmup;
-frontend, encoder and decoding are included, while preparation, file I/O and
-chunk planning are excluded. The audio is JFK's **“We choose to go to the Moon”**
-speech: a **20-second excerpt** and the **18-minute 15-second recording**.
-The quality short-clip test intentionally uses the unmodified five-minute shape.
+frontend, encoder and decoding are included. Preparation, file I/O and chunk
+planning are excluded. Audio is JFK's **“We choose to go to the Moon”** speech:
+a **20-second excerpt** and the **18-minute 15-second recording**.
 
 | Bundle | Audio duration | Transcription | Audio / elapsed |
 | --- | ---: | ---: | ---: |
 | fast | 20 seconds | 0.150 s | 133.4× |
 | fast | 18 min 15 s | 1.876 s | 584.0× |
-| quality | 20 seconds | 1.336 s | 15.0× |
-| quality | 18 min 15 s | 7.044 s | 155.5× |
+| quality | 20 seconds | 0.153 s | 130.9× |
+| quality | 18 min 15 s | 6.257 s | 175.0× |
+
+Quality measurements use three timed runs after warmup, with stable tokens and
+native word timings and nominal thermal state. The full recording uses 74 fast
+chunks or five quality chunks. These contexts and execution policies differ.
 
 Whisper-normalized WER against the supplied long-recording reference:
 
-| Bundle | Word errors | WER |
+| Bundle / reference | Word errors | WER |
 | --- | ---: | ---: |
 | fast | 79/2220 | 3.56% |
-| quality | 44/2220 | 1.98% |
-
-This is one English recording, not a general or multilingual accuracy evaluation.
-Native quantization and FP16 decoder operation order can change token decisions.
-The reported full-recording runs retain all audio, with 74 fast chunks or four
-quality chunks. These are distinct contexts and execution policies, not a
-controlled kernel-only comparison.
+| quality | 45/2220 | 2.03% |
+| FP32 source, same four-minute cuts as quality | 46/2220 | 2.07% |
+
+With the benchmark's simpler normalization, quality scores 54/2219 (2.43%)
+and the matching source 53/2219 (2.39%). This is one English recording, not a
+general accuracy ranking. Quantization and floating-point operation order can
+change token decisions. Both quality shapes pass the short-fixture projected
+encoder check at 2.04% relative RMS versus the independent FP32 source and
+produce identical valid outputs to each other. Word timings are ordered,
+bounded and text-preserving; human word-boundary accuracy was not measured.
+
+A bounded full-recording quality trace contains ANE predictions within all ten
+encoder/subsampling calls and zero target GPU intervals. The decoder uses the
+CPU. Compiler manifests mark both functions in each quality asset fully placed
+on ANE; this is placement evidence, not an arithmetic-utilization measurement.
+Sampled peak client-plus-attributed-neural memory is approximately 1.07 GB,
+excluding unattributed compiler, driver and system memory. Phone execution of
+this quality bundle has not been qualified.
 
 ## Device specialization and loading
 
-Source `.aimodel` files specialize on the device. On the same Mac, preparation
-of these published source assets took:
+Source `.aimodel` files specialize on the device. On the same Mac:
 
 | Bundle | First observed preparation with specialization | Subsequent cached preparation |
 | --- | ---: | ---: |
 | fast | 33 s | 0.060 s |
-| quality | 10 min 7 s | 0.037 s |
+| quality | 6 min 2 s | 0.041 s |
 
-Preparation includes model loading and compilation requested by Core AI, and
-excludes warmup and transcription. These are observed cache histories, not
-guaranteed fresh-install times: the underlying ANE cache state is not fully
-observable, and OS/device/application changes can require specialization again.
+Preparation includes Core AI model initialization and function loading;
+it excludes host sidecar reads, audio I/O, chunk planning, warmup and transcription.
+These are observed cache histories, not guaranteed fresh-install times. The
+underlying ANE cache state is not fully observable, and OS/device/application
+changes can require specialization again.
 
-AoT compiled artifacts
-are omitted because no significant load-time benefit has been demonstrated.
-Authoring debug locations were removed while preserving graph signatures and
-operation counts. `SHA256.json` lists distributed payload checksums.
+AoT compiled artifacts are omitted because no significant load-time benefit has
+been demonstrated. Authoring debug locations were removed while preserving graph
+signatures and operation counts. `SHA256.json` lists distributed payload checksums.

SHA256.json

@@ -11,8 +11,8 @@
   },
   {
     "path": "README.md",
-    "bytes": 4129,
-    "sha256": "e49a5b9f1ea0fa4d4fa8a369f6f9b732da44aae3733b55ea19b696fc384911dc"
+    "bytes": 5082,
+    "sha256": "cf41405ba0c95f42de280675abf725280ad7a682297b3f6d295b839bdb389214"
   },
   {
     "path": "fast/LICENSE",
@@ -137,7 +137,7 @@
   {
     "path": "quality/SHA256.json",
     "bytes": 2155,
-    "sha256": "90a2d9ee273311455b6fb27425c8955cabc7190e0d0bb7cb3f83732d8de80ffb"
+    "sha256": "3a04a0d4d3ff6defb0110aba0e98d678604424bb92b2b18539d06164073eefc3"
   },
   {
     "path": "quality/decoder.f32",
@@ -152,17 +152,17 @@
   {
     "path": "quality/encoder.aimodel/main.hash",
     "bytes": 32,
-    "sha256": "1341561803e3f4ac7781b6dd61f8e1dbb366cd43bcd9ef7a45616f008d29bc5c"
+    "sha256": "ce4c33f50678b578693d3e243d5cfb17015eca0478207ca115f3cb37bef22bd4"
   },
   {
     "path": "quality/encoder.aimodel/main.mlirb",
-    "bytes": 673000984,
-    "sha256": "6d1cf619f908597db03233b283cdee7bc1f0b91af650ffcbb98910a0d7b8e944"
+    "bytes": 672997665,
+    "sha256": "80c2eba65224c1ea94e540310b4150bb9b5f90fc173d22d8e8fbdf46cda54131"
   },
   {
     "path": "quality/encoder.aimodel/metadata.json",
     "bytes": 105,
-    "sha256": "e3632aa1791fdec612c30f4e1c7416d81f346298ea3c3461f79f3a5c79fcec30"
+    "sha256": "bab646772ac556e2994b81d16524116c979df6136b6db03e46024358c11c0a1c"
   },
   {
     "path": "quality/mel-filter.f32",
@@ -171,28 +171,28 @@
   },
   {
     "path": "quality/metadata.json",
-    "bytes": 648,
-    "sha256": "4815b098209eb8e4d5fa5ba882d07da647c809af02e228092ce0fefdcf9f795a"
+    "bytes": 746,
+    "sha256": "ca0164e863a5156c3123d22ef245110c0f4f538482acea013441aee199f0d8d3"
   },
   {
     "path": "quality/runtime.json",
-    "bytes": 222,
-    "sha256": "d0dadc946960eab57e7e2a18777d82de644540ea891737de687f1a48061be01d"
+    "bytes": 224,
+    "sha256": "ce34a23fe78bb828a7c1a2db83a1bac101e8ef0858fee834d218e54a85ceddb8"
   },
   {
     "path": "quality/subsampling.aimodel/main.hash",
     "bytes": 32,
-    "sha256": "0f26d1b91ad6bb0f1d391a6d451ce65189ee0b462181722323c3ffde20b36958"
+    "sha256": "c81c114a265c14e125c95100f939126c5ff31c14cb91d2f04053a1aaa1f968f1"
   },
   {
     "path": "quality/subsampling.aimodel/main.mlirb",
-    "bytes": 4403678,
-    "sha256": "015ca4f66c9f65cd37cf201905f4c4cae7488c2338525327f48f60d294e3fc63"
+    "bytes": 4404378,
+    "sha256": "c36a9e7c2dcc8cb2ecae4675702c7d2a6755a8b18609fec43a9c8905e602a741"
   },
   {
     "path": "quality/subsampling.aimodel/metadata.json",
     "bytes": 105,
-    "sha256": "fb03755e71f71e9b5410a65cce9f492309cd23fcf36d795d55db3f473512cd6c"
+    "sha256": "d27a50da08410c52dc818da0189fee6212906dd78d52732d14a24aca68a5d2f5"
   },
   {
     "path": "quality/vocabulary.json",

quality/SHA256.json

@@ -22,17 +22,17 @@
   {
     "path": "encoder.aimodel/main.hash",
     "bytes": 32,
-    "sha256": "1341561803e3f4ac7781b6dd61f8e1dbb366cd43bcd9ef7a45616f008d29bc5c"
+    "sha256": "ce4c33f50678b578693d3e243d5cfb17015eca0478207ca115f3cb37bef22bd4"
   },
   {
     "path": "encoder.aimodel/main.mlirb",
-    "bytes": 673000984,
-    "sha256": "6d1cf619f908597db03233b283cdee7bc1f0b91af650ffcbb98910a0d7b8e944"
+    "bytes": 672997665,
+    "sha256": "80c2eba65224c1ea94e540310b4150bb9b5f90fc173d22d8e8fbdf46cda54131"
   },
   {
     "path": "encoder.aimodel/metadata.json",
     "bytes": 105,
-    "sha256": "e3632aa1791fdec612c30f4e1c7416d81f346298ea3c3461f79f3a5c79fcec30"
+    "sha256": "bab646772ac556e2994b81d16524116c979df6136b6db03e46024358c11c0a1c"
   },
   {
     "path": "mel-filter.f32",
@@ -41,28 +41,28 @@
   },
   {
     "path": "metadata.json",
-    "bytes": 648,
-    "sha256": "4815b098209eb8e4d5fa5ba882d07da647c809af02e228092ce0fefdcf9f795a"
+    "bytes": 746,
+    "sha256": "ca0164e863a5156c3123d22ef245110c0f4f538482acea013441aee199f0d8d3"
   },
   {
     "path": "runtime.json",
-    "bytes": 222,
-    "sha256": "d0dadc946960eab57e7e2a18777d82de644540ea891737de687f1a48061be01d"
+    "bytes": 224,
+    "sha256": "ce34a23fe78bb828a7c1a2db83a1bac101e8ef0858fee834d218e54a85ceddb8"
   },
   {
     "path": "subsampling.aimodel/main.hash",
     "bytes": 32,
-    "sha256": "0f26d1b91ad6bb0f1d391a6d451ce65189ee0b462181722323c3ffde20b36958"
+    "sha256": "c81c114a265c14e125c95100f939126c5ff31c14cb91d2f04053a1aaa1f968f1"
   },
   {
     "path": "subsampling.aimodel/main.mlirb",
-    "bytes": 4403678,
-    "sha256": "015ca4f66c9f65cd37cf201905f4c4cae7488c2338525327f48f60d294e3fc63"
+    "bytes": 4404378,
+    "sha256": "c36a9e7c2dcc8cb2ecae4675702c7d2a6755a8b18609fec43a9c8905e602a741"
   },
   {
     "path": "subsampling.aimodel/metadata.json",
     "bytes": 105,
-    "sha256": "fb03755e71f71e9b5410a65cce9f492309cd23fcf36d795d55db3f473512cd6c"
+    "sha256": "d27a50da08410c52dc818da0189fee6212906dd78d52732d14a24aca68a5d2f5"
   },
   {
     "path": "vocabulary.json",

quality/encoder.aimodel/main.hash

@@ -1 +1 @@
-m��Y}�23����{���P�˹��׸�D
+���R$���@1AP��_��="����FͥA1

quality/encoder.aimodel/main.mlirb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6d1cf619f908597db03233b283cdee7bc1f0b91af650ffcbb98910a0d7b8e944
-size 673000984
+oid sha256:80c2eba65224c1ea94e540310b4150bb9b5f90fc173d22d8e8fbdf46cda54131
+size 672997665

quality/encoder.aimodel/metadata.json

@@ -1,5 +1,5 @@
 {
-  "creationDate" : "20260926T185446Z",
-  "producer" : "coreai-core 1.0.0b2",
-  "assetVersion" : "2.0"
+  "creationDate" : "20260927T202706Z",
+  "assetVersion" : "2.0",
+  "producer" : "coreai-core 1.0.0b2"
 }

quality/metadata.json

@@ -3,7 +3,8 @@
   "repository": "nvidia/parakeet-tdt-0.6b-v2",
   "revision": "ae9ad07059c7c739ffaf932226a8fe64ae2620b0",
   "frame_buckets": [
-    3776
+    576,
+    3008
   ],
   "precision": "w8a16",
   "blank_id": 1024,
@@ -24,6 +25,9 @@
   "wide_spatial_projections": false,
   "fused_qkv": false,
   "split_glu": true,
+  "native_layer_norm": false,
+  "bounded_norm_variance": false,
+  "accurate_silu": true,
   "subsampling_strategy": "tiled-exact",
   "subsampling_tile_frames": 192
 }

quality/runtime.json

@@ -1,7 +1,7 @@
 {
   "profile": "quality",
   "chunk_policy": "fixed",
-  "chunk_seconds": 300,
+  "chunk_seconds": 240.0,
   "encoder_concurrency": 4,
   "decoder": "cpu",
   "decoder_batch_size": 4,

quality/subsampling.aimodel/main.hash

@@ -1 +1 @@
-
\��l�e�7� ����H�#8RS'�`Ҕ��c
+�j�|-̌��Fup,}*gU���	��:����A

quality/subsampling.aimodel/main.mlirb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:015ca4f66c9f65cd37cf201905f4c4cae7488c2338525327f48f60d294e3fc63
-size 4403678
+oid sha256:c36a9e7c2dcc8cb2ecae4675702c7d2a6755a8b18609fec43a9c8905e602a741
+size 4404378

quality/subsampling.aimodel/metadata.json

@@ -1,5 +1,5 @@
 {
-  "producer" : "coreai-core 1.0.0b2",
+  "creationDate" : "20260927T202709Z",
   "assetVersion" : "2.0",
-  "creationDate" : "20260926T185448Z"
+  "producer" : "coreai-core 1.0.0b2"
 }