quantization_config.json

{
  "model_name": "Qwen/Qwen3-8B",
  "nbits": 4,
  "group_size": 128,
  "tiling_mode": "1D",
  "method": "sinq",
  "compute_dtype": "torch.bfloat16",
  "timestamp": "2025-10-06T17:31:58.512995"
}