autoround-w4/quantization_config.json

{
  "bits": 4,
  "data_type": "int",
  "group_size": 128,
  "sym": true,
  "nsamples": 512,
  "seqlen": 256,
  "autoround_version": "0.13.1",
  "block_name_to_quantize": "model.layers",
  "quant_method": "auto-round",
  "packing_format": "auto_round:auto_gptq"
}