quantization_metadata.json

{
  "model_name": "meta-llama/Llama-3.1-8B",
  "export_timestamp": "2026-03-22T02:50:23.705840",
  "source_run": "/mnt/ssd5/phong/WindQuant_v1/outputs/llama_3.1_8b_20bit/run_20260320_170329",
  "format": "huggingface_safetensors",
  "overall_stats": {
    "average_bitwidth": 1.9216788412573247,
    "compression_ratio": 16.652106123551267,
    "total_original_bits": 223338299392,
    "total_quantized_bits": 13412015137
  }
}