int8

对ONNX模型执行INT8量化，并返回ONNX ModelProto。

函数

quantize

使用编译器友好的启发式方法对ONNX文件应用INT8量化。
quantize(onnx_path, calibration_method='entropy', calibration_data_reader=None, calibration_cache_path=None, calibration_shapes=None, calibration_eps=['cuda:0', 'cpu', 'trt'], op_types_to_quantize=None, op_types_to_exclude=None, nodes_to_quantize=None, nodes_to_exclude=None, use_external_data_format=True, intermediate_generated_files=[], verbose=False, trt_extra_plugin_lib_paths=None, high_precision_dtype='fp32', **kwargs)

使用编译器友好的启发式方法对ONNX文件应用INT8量化。

支持对['Add', 'AveragePool', 'BatchNormalization', 'Clip', 'Conv', 'ConvTranspose', 'Gemm', 'GlobalAveragePool', 'MatMul', 'MaxPool', 'Mul']操作类型进行量化。

Parameters:

  • onnx_path (str) –

  • calibration_method (str) –

  • calibration_data_reader (CalibrationDataReader) –

  • calibration_cache_path (str) –

  • calibration_shapes (str) –

  • calibration_eps (List[str]) –

  • op_types_to_quantize (List[str]) –

  • op_types_to_exclude (List[str]) –

  • nodes_to_quantize (List[str]) –

  • nodes_to_exclude (List[str]) –

  • use_external_data_format (bool) –

  • intermediate_generated_files (List[str]) –

  • verbose (bool) –

  • trt_extra_plugin_lib_paths (str) –

  • high_precision_dtype (str) –

Return type:

ModelProto