{
  "key": "nvfp4",
  "mode": "nvfp4",
  "bits": 4,
  "group_size": 16,
  "source_repo": "Accio-Lab/occamy-1.0",
  "source_revision": "8f8e0e58a3c9df042be1a3fa2c191fd8047acfb8",
  "source_payload_sha256_rechecked": true,
  "requantized": false,
  "conversion_backend": "Linux native MLX CUDA 12",
  "router_shared_gate_mode": "affine",
  "router_shared_gate_bits": 8,
  "router_shared_gate_group_size": 64,
  "versions": {
    "mlx": "0.32.2",
    "mlx-lm": "0.31.3",
    "transformers": "5.8.1",
    "numpy": "2.5.3"
  },
  "layout_preflight": {
    "passed": true,
    "source_index_tensors": 31666,
    "layers": 40,
    "experts_per_layer": 256,
    "groups": 120,
    "independent_expert_tensors": 30720,
    "all_shapes_checked": true,
    "each_group_dtype_uniform": true,
    "weight_payload_read": false,
    "model_instantiated": false
  },
  "layout_adapter": {
    "calls": [
      {
        "adapted": true,
        "groups": 120,
        "source_expert_tensors": 30720,
        "expert_order": "numeric 0..num_experts-1",
        "value_transform": "stack only; no cast, quantization, norm or conv transform"
      }
    ],
    "native_sanitize_calls": 1,
    "restored": true
  },
  "seconds": 42.45400047302246,
  "candidate_only": true,
  "mac_acceptance": "pending",
  "vision_included": false,
  "mtp_included": false
}