WarpStateAI / model /settings.json
zerdovzad's picture
Upload 8 files
e5e74c7 verified
Raw History Blame Contribute Delete
4.28 kB
{
"signature": {
"plan": {
"format": "warpstate-posttrain-plan-v1",
"stage": "sft",
"after_completion": "pause_for_evaluation",
"base": {
"step": 76294,
"tokens_seen": 10000007168,
"tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e"
},
"handoff": {
"base_work_dir": "rtx6000_40b/work",
"base_program": "warpstate-train",
"supervisor_socket": "rtx6000_40b/work/control/supervisor.sock",
"archive_dir": "rtx6000_40b/work/archives/step-00076294",
"work_dir": "rtx6000_40b/work/posttrain/step-00076294",
"data_dir": "rtx6000_40b/work/posttrain/data",
"tokenizer": "rtx6000_40b/tokenizer.model",
"processes": 2,
"poll_seconds": 15
},
"training": {
"sequence_length": 2048,
"micro_batch": 4,
"grad_accum": 4,
"lr": 5e-06,
"min_lr": 0.0,
"warmup_ratio": 0.03,
"epochs": 1,
"max_steps": null,
"weight_decay": 0.0,
"seed": 42,
"eval_every": 100,
"save_every": 100,
"compile_core": true,
"loss_mode": "compiled-full",
"bf16_residuals": true,
"gradient_checkpointing": false,
"workers": 8
}
},
"settings": {
"sequence_length": 2048,
"micro_batch": 4,
"grad_accum": 4,
"lr": 5e-06,
"min_lr": 0.0,
"warmup_ratio": 0.03,
"epochs": 1,
"max_steps": null,
"weight_decay": 0.0,
"seed": 42,
"eval_every": 100,
"save_every": 100,
"eval_batches": 32,
"compile_core": true,
"loss_mode": "compiled-full",
"bf16_residuals": true,
"gradient_checkpointing": false,
"workers": 8,
"deterministic": true,
"clip_grad_norm": 1.0
},
"data_sha256": "70b144f08930fa2a6b4fde19bb26b8dee4c90bd6fe464cfb23282e7eb986801e",
"tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e",
"renderer": "warpstate-segmented-roles-v1",
"base_sha256": "dc7605cecd6c807135e3d83e04496fa96079fef022bed1952c8cf21df5f573da",
"world_size": 2,
"device_type": "cuda",
"torch_version": "2.14.0+cu130",
"sources": {
"posttrain.py": "bd10e97e61b3289e83b7e79c76042bf7f04ce58f4af9067db97ac1fe60d4a411",
"ops/posttrain_data.py": "486d2e96e7e99ffd05caf0de7d92068ce47047d4641ede3515681cbb79a9ce10",
"ops/loss_experiment.py": "12bf7822979692d5106e30544d0ed5d977fb06f556c068ef5f92ddf56ab8376e",
"ops/residual_experiment.py": "e82aa68e36c76a228156016eb04a02ce1a632a63ac6eda567ec9c4d258aae8f1",
"ops/checkpoint_writer.py": "3c7fe494e8782802b93a439c05ddd6f416711ca1b3f0f8a196a97a48e37f0216",
"run_40b.py": "9ad6533d781e376b3b40c16e6c29ef15063d39f917a93800489f17115be15ba0",
"warpstate/__init__.py": "f386c0880cb499d84159d8205667d250a72d0e8a102f0d7be6af0ba0ea33b22b",
"warpstate/config.py": "1d716b6c582bab21e8d5fcbaa6d9f3be4f792baf9b6abe826c72422f8eced6ad",
"warpstate/model.py": "025330ee58cca899695c7631e1a06d2efd006402c61e51106afe6fa2e1d35161"
}
},
"origin": {
"format": "warpstate-chat-v1",
"inference_only": true,
"config": {
"vocab_size": 32768,
"d_model": 2048,
"n_heads": 32,
"n_cores": 28,
"logical_depth": 28,
"ffn_hidden": 7168,
"chunk_size": 128,
"dropout": 0.0,
"rms_eps": 1e-05,
"gradient_checkpointing": false,
"parallel_memory": true,
"loss_chunk_size": 512
},
"step": 76294,
"tokens_seen": 10000007168,
"val_loss": 2.3496886491775513,
"tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e",
"captured_at_utc": "2026-09-14T14:27:16Z",
"source": {
"path": "/workspace/WarpState/rtx6000_40b/work/run/latest.pt",
"device": 66,
"inode": 8616307015,
"size": 24061705357,
"mtime_ns": 1789396019640599046
},
"weights_sha256": "dc7605cecd6c807135e3d83e04496fa96079fef022bed1952c8cf21df5f573da",
"weights_path": "/workspace/WarpState/rtx6000_40b/work/posttrain/step-00076294/initialization/model.pt"
},
"total_steps": 10312,
"train_rows": 329973,
"val_rows": 1603
}