quantize_preprocess

Description

Preprocesses the quantization of a graph based on the quantization configuration file, inserts the balanced quantization operators, generates a balanced-factor record file record_file, and returns an ONNX model ready for calibration.

Prototype

quantize_preprocess(config_file, record_file, model_file, modified_onnx_file)

Command-Line Options

Option	Input/Return	Description	Restriction
config_file	Input	User-defined quantization configuration file, which specifies the configuration of each layer to be quantized.	A string
record_file	Input	Directory of the balanced-factor record file, including the file name.	A string
model_file	Input	Source ONNX model file or updated ONNX model file generated by create_quant_config.	A string
modified_onnx_file	Input	File name of the ONNX model ready for calibration, whose activations are to be quantized.	A string

Return Value

None.

Outputs

None

Examples

import amct_onnx as amct

model_file = "resnet101.onnx"
tensor_balance_factor_record_file  = os.path.join(TMP, 'tensor_balance_factor_record.txt')
modified_model = os.path.join(TMP, 'modified_model.onnx')
config_file="./configs/config.json"
# Insert the quantization API.
amct.quantize_preprocess(config_file,
                         tensor_balance_factor_record_file,
                         model_file,
                         modified_model)

Parent topic: PTQ APIs