quantization_config.json

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