{ "data_sha256": "70b144f08930fa2a6b4fde19bb26b8dee4c90bd6fe464cfb23282e7eb986801e", "format": "warpstate-sft-archive-v1", "model_sha256": "ddf25c0260a189614613c4ceb3b778983aec601c55ff4a175aff350d85ce1f44", "raw_base_sha256": "dc7605cecd6c807135e3d83e04496fa96079fef022bed1952c8cf21df5f573da", "raw_base_step": 76294, "raw_base_tokens": 10000007168, "repository_revision": "774890d7f88c31fbe3e417aebab45ab08a177208", "sft_assistant_tokens": 143695039, "sft_input_tokens": 212321372, "sft_signature": { "base_sha256": "dc7605cecd6c807135e3d83e04496fa96079fef022bed1952c8cf21df5f573da", "data_sha256": "70b144f08930fa2a6b4fde19bb26b8dee4c90bd6fe464cfb23282e7eb986801e", "device_type": "cuda", "plan": { "after_completion": "pause_for_evaluation", "base": { "step": 76294, "tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e", "tokens_seen": 10000007168 }, "format": "warpstate-posttrain-plan-v1", "handoff": { "archive_dir": "rtx6000_40b/work/archives/step-00076294", "base_program": "warpstate-train", "base_work_dir": "rtx6000_40b/work", "data_dir": "rtx6000_40b/work/posttrain/data", "poll_seconds": 15, "processes": 2, "supervisor_socket": "rtx6000_40b/work/control/supervisor.sock", "tokenizer": "rtx6000_40b/tokenizer.model", "work_dir": "rtx6000_40b/work/posttrain/step-00076294" }, "stage": "sft", "training": { "bf16_residuals": true, "compile_core": true, "epochs": 1, "eval_every": 100, "grad_accum": 4, "gradient_checkpointing": false, "loss_mode": "compiled-full", "lr": 5e-06, "max_steps": null, "micro_batch": 4, "min_lr": 0.0, "save_every": 100, "seed": 42, "sequence_length": 2048, "warmup_ratio": 0.03, "weight_decay": 0.0, "workers": 8 } }, "renderer": "warpstate-segmented-roles-v1", "settings": { "bf16_residuals": true, "clip_grad_norm": 1.0, "compile_core": true, "deterministic": true, "epochs": 1, "eval_batches": 32, "eval_every": 100, "grad_accum": 4, "gradient_checkpointing": false, "loss_mode": "compiled-full", "lr": 5e-06, "max_steps": null, "micro_batch": 4, "min_lr": 0.0, "save_every": 100, "seed": 42, "sequence_length": 2048, "warmup_ratio": 0.03, "weight_decay": 0.0, "workers": 8 }, "sources": { "ops/checkpoint_writer.py": "3c7fe494e8782802b93a439c05ddd6f416711ca1b3f0f8a196a97a48e37f0216", "ops/loss_experiment.py": "12bf7822979692d5106e30544d0ed5d977fb06f556c068ef5f92ddf56ab8376e", "ops/posttrain_data.py": "486d2e96e7e99ffd05caf0de7d92068ce47047d4641ede3515681cbb79a9ce10", "ops/residual_experiment.py": "e82aa68e36c76a228156016eb04a02ce1a632a63ac6eda567ec9c4d258aae8f1", "posttrain.py": "bd10e97e61b3289e83b7e79c76042bf7f04ce58f4af9067db97ac1fe60d4a411", "run_40b.py": "9ad6533d781e376b3b40c16e6c29ef15063d39f917a93800489f17115be15ba0", "warpstate/__init__.py": "f386c0880cb499d84159d8205667d250a72d0e8a102f0d7be6af0ba0ea33b22b", "warpstate/config.py": "1d716b6c582bab21e8d5fcbaa6d9f3be4f792baf9b6abe826c72422f8eced6ad", "warpstate/model.py": "025330ee58cca899695c7631e1a06d2efd006402c61e51106afe6fa2e1d35161" }, "tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e", "torch_version": "2.14.0+cu130", "world_size": 2 }, "sft_step": 10312, "source_checkpoint_sha256": "b94426834e5c56895ed34529479fe540ff14c2b6cef9c6804b9cf16a34ec27c8", "tokenizer_sha256": "ebae4b4310068ee102f7bf68ac68b79ee1425baa32de3ca8ff8138692529d84e" }