{
  "attention_projections_full_precision": true,
  "calibration": {
    "calls": 3934,
    "conversations": 3,
    "effective_rows_lower_bound": 3934,
    "input": "captured exact combined_embeds; no tokenization",
    "manifest_sha256": "8522310fb27db4ebe075628b720058693377a8cc902e1155187f38e2bd02efb8",
    "selection_manifest_sha256": "4d38a0f3384a2f0695a60f42ad22c5945be6c0b094d5628779ca1776f4ad3402"
  },
  "embeddings_full_precision": true,
  "evaluation": {
    "calls": 1172,
    "conversations": 1,
    "disjoint_from_calibration": true,
    "manifest_sha256": "2080deb829fcb9cd6fc524edc0dca86cb591af1d9d8abfd10dbc9632e15b1a6b",
    "selection_manifest_sha256": "4d38a0f3384a2f0695a60f42ad22c5945be6c0b094d5628779ca1776f4ad3402"
  },
  "function_head_full_precision": true,
  "kind": "portable_nano_conversion_report",
  "layer_metrics": [],
  "method": "calibrated symmetric GPTQ MLP+Mamba+text-head W8A16 group-128; K=15680 MLP-down padded to K=15744; function head and embeddings full precision",
  "policy": "w8-all-g128",
  "quantization_config": {
    "bits": 8,
    "checkpoint_format": "gptq",
    "desc_act": false,
    "dynamic": {
      "-:.*(?:embed_tokens|embed_asr_tokens)$": {},
      "-:.*(?:q_proj|k_proj|v_proj|o_proj)$": {},
      "-:.*function_head$": {}
    },
    "group_size": 128,
    "lm_head": true,
    "modules_in_block_to_quantize": [
      "mixer.up_proj",
      "mixer.down_proj",
      "mixer.in_proj",
      "mixer.out_proj",
      "lm_head"
    ],
    "quant_method": "gptq",
    "sym": true
  },
  "quantized_output_heads": [
    "lm_head"
  ],
  "reproducibility": {
    "byte_identical": true,
    "runs": 2,
    "tree_sha256": "0d7c0d8e8503ece71b2b606c86dad371c3d7fc5b58d892945068b14331029bb3"
  },
  "runtime": {
    "driver_version": "580.142",
    "kernel_version": "6.17.0-1014-nvidia",
    "packages": {
      "compressed-tensors": "0.13.0",
      "nvidia-modelopt": "0.37.0",
      "safetensors": "0.8.0",
      "torch": "2.10.0a0+b4e4ee81d3.nv25.12",
      "transformers": "4.56.0",
      "vllm": "0.17.1.dev0+gb31e9326a.fi065"
    }
  },
  "schema": 1,
  "selected_source_bytes": 32314490880,
  "selected_tensor_count": 105,
  "shape_exception": {
    "effective_group_size": 128,
    "other_group_size_fallbacks_allowed": false,
    "padded_shape": [
      4480,
      15744
    ],
    "reason": "zero-pad K by 64 so all 105 selected tensors remain group-128",
    "requested_group_size": 128,
    "runtime_padding": true,
    "selector": "stt_model.llm.layers.*.mixer.down_proj.weight",
    "shape": [
      4480,
      15680
    ]
  },
  "source_provenance": {
    "checkpoint_sha256": "d553750c29434a6bb524377e17634c6cafdbf621892e643a77f406e51570354b",
    "derived_from_ea": false,
    "nano_sha256": "1b473452103362817e9408cf864aa8ef8ca92a48ef8fd94cb3d40fa7cf14e44b",
    "repository": "nvidia/NVIDIA-NemotronLabs-VoiceChat-11B",
    "revision": "fb0f94eaf4d03ddc430f39565229393fa1b50c26"
  },
  "status": "complete"
}