quantization_config.json

{
  "bits": 4,
  "data_type": "int",
  "group_size": 128,
  "sym": true,
  "batch_size": 1,
  "gradient_accumulate_steps": 8,
  "seqlen": 512,
  "autoround_version": "0.13.0",
  "block_name_to_quantize": "model.layers",
  "quant_method": "auto-round",
  "packing_format": "auto_round:auto_gptq",
  "extra_config": {
    "model.layers.0.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.1.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.2.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.3.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.4.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.5.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.6.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.7.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.8.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.9.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.10.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.11.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.12.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    "model.layers.13.mlp.router.classifier": {
      "bits": 16,
      "data_type": "float"
    },
    ".*classifier.*": {
      "bits": 16,
      "data_type": "float"
    }
  }
}