quantization_config.json

{
  "bits": 3,
  "data_type": "int",
  "group_size": 128,
  "sym": false,
  "iters": 300,
  "seqlen": 256,
  "autoround_version": "0.10.2",
  "quant_method": "auto-round",
  "packing_format": "auto_round"
}