mctd-checkpoints / resources.json
oosnow's picture
Attach completed PlanCraft SFT full-test evaluation: 112/580
91af209 verified
Raw
History Blame Contribute Delete
6.05 kB
{
"format_version": 1,
"packages": [
{
"file": "packs/initialization.tar.gz",
"sha256": "355f5eee9b33efd013969674e5a2f00db18d2123996089332abc8b35892e4795",
"bytes": 9787343,
"environment": null,
"files": {
"resources/wa_init.pt": {
"sha256": "340edef6e65fcfe05bab75f8be2214ad11df49b6ad48544df7f9c51dfe424396",
"bytes": 16778648
},
"resources/action_head.pt": {
"sha256": "f274be89e3f7c739c2da6cae40e0d997c44d469a7afd469246ba599f61f72daf",
"bytes": 10516872
}
},
"summary": {
"action_head_initialization": "historical expert-SFT warm start; not random",
"wa_initialization": "existing student-dependent initialized Wa"
}
},
{
"file": "packs/alfworld-latent-sft-k4.tar.gz",
"sha256": "e0bbd26fbc633e7391b0cd7bec2489fe072794f9b1068c22a543e94dc53d6459",
"bytes": 9794834,
"environment": "alfworld",
"files": {
"experiments/pretrained/alfworld/latent-sft-k4/evaluation_spec.json": {
"sha256": "ffaadeee0b05556aec4141ce8bf7ec26ef495b40cbd0569e85cb59e9fbd2aee8",
"bytes": 331
},
"experiments/pretrained/alfworld/latent-sft-k4/manifest.json": {
"sha256": "781af5501be663bd053eb693a363d261314fd8ec240cc71158560e4e49d80924",
"bytes": 250
},
"experiments/pretrained/alfworld/latent-sft-k4/resolved_config.json": {
"sha256": "8c443712976c54cdab8ef83de9d2cb0bbf7f18a3036589b0006c0fc9b779abf6",
"bytes": 267
},
"experiments/pretrained/alfworld/latent-sft-k4/artifacts/wa.pt": {
"sha256": "a334a934823d7d243e9c5fe122aabd782f5b1bec501faa5394778e189b3c2f99",
"bytes": 16778870
},
"experiments/pretrained/alfworld/latent-sft-k4/artifacts/action_head.pt": {
"sha256": "6b705372f89aa7462be395fb5abbd07bffa7024dcabec60fa0e0245d3d49096e",
"bytes": 10517383
},
"experiments/pretrained/alfworld/latent-sft-k4/report.json": {
"sha256": "0cc50f85b907a16310a80ef478fdf71d994bd4b12cc5e7f42d69d7a1ce607700",
"bytes": 136
}
},
"summary": {
"use": "OPD initialization and evaluation, not optimizer-state resume",
"source_run": "latent-sft--qwen3-30b-a3b-instruct--action-head--k4--20260819-112619-BJT"
}
},
{
"file": "packs/eb-habitat-initialization-20260909.tar.gz",
"bytes": 3863820,
"sha256": "a87f0309d40a2e1379c4d01f6d74481f80363744226c39fbebc6f23e7dbc9dca",
"environment": "eb-habitat",
"files": {
"resources/eb-habitat-action-head.pt": {
"bytes": 4222551,
"sha256": "d7da6fcfe487910f782e921691a6ea5c296146985157a04d11ecc28e1448c3a6"
}
},
"summary": {
"initialization": "seeded random untrained action head",
"seed": 42,
"student": "Qwen/Qwen3-1.7B",
"not_a_trained_checkpoint": true
}
},
{
"file": "packs/plancraft-latent-sft-k8-20260910.tar.gz",
"environment": "plancraft",
"bytes": 13824385,
"sha256": "56b63464e89bb8c618ca44ce0b045b41678c4cfbefdf28dd460172c88f250f24",
"files": {
"experiments/pretrained/plancraft/latent-sft-k8/artifacts/action_head.pt": {
"bytes": 10518121,
"sha256": "d157be5a3f7d39bed35253c472ef3e23ba2af12b3701ddef598ac000bf711e3a"
},
"experiments/pretrained/plancraft/latent-sft-k8/artifacts/wa.pt": {
"bytes": 16778648,
"sha256": "340edef6e65fcfe05bab75f8be2214ad11df49b6ad48544df7f9c51dfe424396"
},
"experiments/pretrained/plancraft/latent-sft-k8/evaluation_spec.json": {
"bytes": 333,
"sha256": "2bdcd0052c73206de01d71cda233c402ac824cd546c3becba5cc619d38c3b86e"
},
"experiments/pretrained/plancraft/latent-sft-k8/evaluations/full-test-pass1/report.json": {
"bytes": 4500,
"sha256": "f6bca55446d6d9b657f874b20d24405a5293daa3148f6368c0e4c3da7881a0d1"
},
"experiments/pretrained/plancraft/latent-sft-k8/evaluations/full-test-pass1/resolved_config.json": {
"bytes": 2833,
"sha256": "e9df7ffc5df0c2cdb813356dbe9f1599fb02e3672bc22a84010b86b967829a46"
},
"experiments/pretrained/plancraft/latent-sft-k8/manifest.json": {
"bytes": 314,
"sha256": "964ae625e59e80bf7d7f3b34c78fe57b39cd5815772fd933d49f7d660b89aded"
},
"experiments/pretrained/plancraft/latent-sft-k8/report.json": {
"bytes": 87,
"sha256": "98b149743be21f73fca4a61c3c7e13e865fc45ae9ae5af303b4f6a73377a02e6"
},
"experiments/pretrained/plancraft/latent-sft-k8/resolved_config.json": {
"bytes": 269,
"sha256": "d04b52bce5fc4e39d616c4329665d2558deb2f0f996e831a48dff095372a17fb"
},
"experiments/pretrained/plancraft/latent-sft-k8/source_training_config.json": {
"bytes": 5084,
"sha256": "c9f62721c281c4f8d764b1d289265679ce47632194664ad2d3d8b1f255e5819e"
},
"data/pretrained/plancraft/sft-20260910/commands/command_embeddings.safetensors": {
"bytes": 5152856,
"sha256": "625d4cb7296ba37c5a230530a8424235cd45e04e354e1790022d9042be6f5113"
},
"data/pretrained/plancraft/sft-20260910/commands/command_vocab.json": {
"bytes": 57910,
"sha256": "a79ecf5ed77ba85ddb6f9981a7e0e7e66592c2f792747caada0cdbe4d8a08cbb"
},
"data/pretrained/plancraft/sft-20260910/commands/manifest.json": {
"bytes": 595,
"sha256": "e8a097b9fb9ac25add7c47d6247d7a3441b249c85ddcf14a40323c27df24443e"
}
},
"summary": {
"source_run": "20260910-154609-BJT--latent-sft--plancraft--k8--m1",
"checkpoint": "best",
"base_model": "Qwen/Qwen3-1.7B",
"latent_steps": 8,
"full_test_tasks": 580,
"success_rate": 0.19310344827586207
}
}
]
}