coding-router v6: ONNX (int8+fp32) + PyTorch weights, tokenizer, calibration, eval results
e07ba76 verified | { | |
| "method": "per_head_temperature_scaling", | |
| "source_split": "validation", | |
| "per_head": { | |
| "complexity": 0.891251, | |
| "task_type": 0.707946, | |
| "risk": 1.059254 | |
| } | |
| } |