| { |
| "artifact": { |
| "intended_repo": "WaveCut/Qwen3.8-27B-CRACK-DWQ-4bit-MLX", |
| "library": "mlx", |
| "pipeline_tag": "image-text-to-text" |
| }, |
| "build": { |
| "repository_commit": "e7a7155", |
| "method": "DWQ sparse-teacher distillation of affine scales and biases", |
| "teacher": "pinned source MXFP8 CRACK", |
| "temperature": 2.0, |
| "learning_rate": 3e-7, |
| "train_updates": 128, |
| "heldout_records": 16, |
| "heldout_loss": { |
| "rtn_q4_g64": 0.168036, |
| "dwq_q4_g64": 0.093706, |
| "relative_reduction": 0.44235 |
| } |
| }, |
| "calibration": { |
| "unique_records": 128, |
| "mix": { |
| "tool_calling": 48, |
| "swe_agent": 32, |
| "multilingual": 32, |
| "code": 16 |
| }, |
| "datasets": [ |
| "NousResearch/hermes-function-calling-v1", |
| "nebius/SWE-agent-trajectories", |
| "CohereLabs/aya_dataset", |
| "openai/openai_humaneval" |
| ], |
| "crc32": { |
| "source_manifest": "ed8b9ebf", |
| "pilot_train_48": "c137585c", |
| "continuation_train_80": "7e7f5e86", |
| "heldout_16": "78835208" |
| } |
| }, |
| "components": { |
| "language": { |
| "tensors": 1847, |
| "quantization": { |
| "bits": 4, |
| "group_size": 64, |
| "mode": "affine" |
| } |
| }, |
| "vision": { |
| "tensors": 417, |
| "value_identical_to_source": true |
| }, |
| "mtp": { |
| "tensors": 31, |
| "byte_identical_to_source": true, |
| "shard_crc32": "d54df83d", |
| "quantization": { |
| "bits": 8, |
| "group_size": 128, |
| "mode": "affine" |
| } |
| }, |
| "chat_template": { |
| "byte_identical_to_source": true, |
| "crc32": "367cbad5" |
| } |
| }, |
| "runtime": { |
| "platform": "Apple Silicon Metal", |
| "mlx": "0.32.0", |
| "mlx_lm": "0.31.3", |
| "mlx_vlm": "0.6.13", |
| "runtime_fork_required": false, |
| "custom_model_code_required": false |
| }, |
| "source": { |
| "repo": "dealignai/Qwen3.8-27B-MXFP8-CRACK", |
| "revision": "7d2f222cbc732a35d3be65710c6053a3b28e1493" |
| } |
| } |
|
|
