| { |
| "repo_id": "FlyinGodzilla/AMOR-Mamba2-440M", |
| "source_filename": "amor_440m_step_0018255.pt", |
| "source_sha256": "9a2e1000be0ab7c2bc5003e5a8f6ca4bf5e357828346f4aa19860fffa6c901df", |
| "step": 18255, |
| "tokens_seen": 8973017088, |
| "n_params": 448650496, |
| "tensor_dtypes": { |
| "torch.float32": 338 |
| }, |
| "gate_buffers": { |
| "0": { |
| "running_threshold": 0.8732627034187317, |
| "running_std": 0.046764787286520004 |
| }, |
| "1": { |
| "running_threshold": 0.8569177389144897, |
| "running_std": 0.05602722615003586 |
| }, |
| "2": { |
| "running_threshold": 0.8578681349754333, |
| "running_std": 0.05078088119626045 |
| } |
| }, |
| "optimizer_included": false, |
| "precision_changed": false, |
| "validation": { |
| "exported_tensors_equal": true, |
| "strict_model_load": true, |
| "generation": true, |
| "device": "cpu", |
| "dtype": "float32", |
| "torch": "2.6.0", |
| "transformers": "4.51.3", |
| "tokenizer_revision": "5590cc63198b0f111f4289daf006da0a12044173", |
| "prompt": "The Eiffel Tower is", |
| "generated_token_ids": [ |
| 279, |
| 82717, |
| 4857, |
| 304 |
| ], |
| "prefill_max_abs_diff": 0.0, |
| "decode_max_abs_diff": 2.288818359375e-05, |
| "seconds": 51.611721992492676, |
| "cuda": { |
| "versions": { |
| "torch": "2.6.0+cu124", |
| "transformers": "5.0.0", |
| "mamba-ssm": "2.3.0", |
| "causal-conv1d": "1.6.0", |
| "flash-linear-attention": "0.4.2", |
| "triton": "3.2.0", |
| "safetensors": "0.7.0" |
| }, |
| "tokenizer_revision": "5590cc63198b0f111f4289daf006da0a12044173", |
| "strict_model_load": true, |
| "generation": true, |
| "device": "NVIDIA H100 NVL", |
| "torch": "2.6.0+cu124", |
| "autocast": "bfloat16", |
| "prefill_max_abs_diff": 0.0, |
| "decode_max_abs_diff": 0.0625, |
| "generated_token_ids": [ |
| 264, |
| 734, |
| 430, |
| 4780, |
| 279, |
| 1121, |
| 315, |
| 264 |
| ], |
| "published_cli_passed": true, |
| "test_scope": "Short CUDA smoke test; not a benchmark rerun", |
| "comparison_tolerance": { |
| "rtol": 0.02, |
| "atol": 0.1 |
| }, |
| "tested_decoder_sha256": "4b48db8e5fb67318244c533356903f40703156cdca24ccf4fd49086e72b16db7" |
| }, |
| "router_release": { |
| "exact_router_export": true, |
| "base_model_association": true, |
| "default_forward_max_abs_diff": 0.0, |
| "modes": [ |
| { |
| "router": false, |
| "cases": [ |
| { |
| "batch": 1, |
| "length": 16, |
| "prefill_max_abs_diff": 0.0, |
| "decode_max_abs_diff": 0.0625 |
| }, |
| { |
| "batch": 1, |
| "length": 128, |
| "prefill_max_abs_diff": 0.0625, |
| "decode_max_abs_diff": 0.09375 |
| }, |
| { |
| "batch": 2, |
| "length": 64, |
| "prefill_max_abs_diff": 0.0 |
| } |
| ], |
| "canonical_router_exact": null |
| }, |
| { |
| "router": true, |
| "cases": [ |
| { |
| "batch": 1, |
| "length": 16, |
| "prefill_max_abs_diff": 0.0625, |
| "decode_max_abs_diff": 0.0625 |
| }, |
| { |
| "batch": 1, |
| "length": 128, |
| "prefill_max_abs_diff": 0.0, |
| "decode_max_abs_diff": 0.09375 |
| }, |
| { |
| "batch": 2, |
| "length": 64, |
| "prefill_max_abs_diff": 0.0 |
| } |
| ], |
| "canonical_router_exact": true |
| } |
| ], |
| "frozen_thresholds_preserved": true, |
| "training_router_rejected": true, |
| "gpu": "NVIDIA H100 NVL", |
| "torch": "2.6.0+cu124", |
| "generation_cli_modes_passed": [ |
| "entropy", |
| "router" |
| ] |
| }, |
| "sharded_release": { |
| "all_tensor_bytes_identical": true, |
| "strict_model_load": true, |
| "entropy_and_router_generation_passed": true, |
| "unit_tests_passed": 17, |
| "device": "NVIDIA H100 NVL" |
| }, |
| "current_release": { |
| "strict_model_load": true, |
| "entropy_and_router_generation_passed": true, |
| "unit_tests_passed": 17, |
| "device": "NVIDIA H100 NVL" |
| } |
| }, |
| "implementation": "AMOR shared release: native entropy gate and optional fitted inference router", |
| "router": { |
| "variant": "distill", |
| "hidden_dim": 512, |
| "activation": "silu", |
| "weights": "router.safetensors", |
| "config": "router_config.json", |
| "sha256": "4ccfb19021cfe868dd0d5fdc5bbd32274f2be51ebdcafcf75601cb102c030fc2", |
| "base_weight_sha256": "75a45229eb09553f69cf14b69fb368e7d1fb3296a1e146002f6f9ed0ff650f47", |
| "precision": "float32", |
| "inference_only": true |
| }, |
| "weight_layout": { |
| "format": "safetensors-sharded-v1", |
| "index": "model.safetensors.index.json", |
| "shards": 3, |
| "weight_identity_sha256": "75a45229eb09553f69cf14b69fb368e7d1fb3296a1e146002f6f9ed0ff650f47", |
| "original_single_file_sha256": "3248a9439b813c8c15219509079562380d10a82fcadf88c78c79cb9694fa0339", |
| "all_tensor_bytes_identical": true |
| } |
| } |
|
|
