q4/quantize_config.json

{
  "quantization": {
    "bits": 4,
    "group_size": 64
  }
}