onnx/quantize_config.json

{
  "quantization": {
    "int8": "onnxruntime dynamic quantization, QInt8",
    "uint8": "onnxruntime dynamic quantization, QUInt8",
    "quantized": "copy of int8 for compatibility",
    "q4": "MatMulNBitsQuantizer, bits=4, block_size=32, symmetric"
  },
  "export_tool": "optimum-cli export onnx",
  "export_task": "image-classification",
  "onnx_version": "v1.1"
}