quantization_config.json

{
  "add_skip_keys": false,
  "dequantize_fp32": false,
  "group_size": 0,
  "is_integer": true,
  "modules_dtype_dict": {},
  "modules_to_not_convert": [],
  "non_blocking": false,
  "quant_conv": false,
  "quant_method": "sdnq",
  "quantization_device": null,
  "return_device": null,
  "svd_rank": 32,
  "svd_steps": 8,
  "use_quantized_matmul": false,
  "use_quantized_matmul_conv": false,
  "use_svd": true,
  "weights_dtype": "uint4"
}