quantization_config.json

{
  "bits": 4,
  "group_size": 128,
  "sym": true,
  "data_type": "int",
  "iters": 1000,
  "nsamples": 512,
  "low_gpu_mem_usage": true,
  "autoround_version": "0.7.1",
  "quant_method": "auto-round",
  "packing_format": "auto_round:auto_gptq"
}