{
  "quantization": {
    "group_size": 64,
    "bits": 6
  },
  "precision": "int6"
}