quantization_config.json

{
  "bits": 4,
  "group_size": 128,
  "sym": true,
  "data_type": "int",
  "seqlen": 128,
  "batch_size": 1,
  "lr": 0.005,
  "minmax_lr": 0.005,
  "iters": 50,
  "amp": false,
  "autoround_version": "0.7.1",
  "quant_method": "auto-round",
  "packing_format": "auto_round:auto_gptq"
}