{
  "source_revision": "60d858b518b4e19d404af3737f848fc185b30177",
  "source_safetensors_bytes": 17348198658,
  "output_safetensors_bytes": 7000335408,
  "compression_ratio": 2.478195350207711,
  "quantized_projections": 196,
  "method": "Activation-aware AWQ using llm-compressor; exact layout repack into AWQ GEMM",
  "lm_head": "BF16",
  "scheme": "W4A16_ASYM",
  "group_size": 128,
  "duo_scaling": "both",
  "grid_points": 40,
  "calibration_samples": 256,
  "audio_calibration_samples": 128,
  "text_calibration_samples": 128,
  "calibration_stats": {
    "audio_sessions": 128,
    "audio_rows_mean": 944.7734375,
    "audio_rows_total": 120931,
    "truncated_at_2048": 0
  },
  "max_sequence_length": 2048,
  "packing_validation": "All integer weight and zero-point codes round-trip exactly; all scale values preserved exactly",
  "pruning": {
    "removed_prefixes": [
      "model.acoustic_tokenizer.decoder."
    ],
    "removed_bf16_bytes": 687391938,
    "speech_bytes": 1429418752
  }
}