perf_based_auto_calibration

Applicable Products

Product	Supported
Ascend 950PR/Ascend 950DT	x
Atlas A3 training products/Atlas A3 inference products	x
Atlas A2 training products/Atlas A2 inference products	x
Atlas 200I/500 A2 inference products	x
Atlas inference products	x
Atlas training products	x

Description

Searches for a mixed-precision quantized model with better performance based on the original dequantized model, quantized model, and performance sampler configuration file input by the user, and outputs a fake-quantized model that can be used for accuracy simulation in the TensorFlow environment and a deployable model that can be used for inference on the AI processor.

Prototype

perf_based_auto_calibration(original_model_file, quantize_model_file, outputs, sampler_config_file, save_dir, strategy='BatchRollBack')

Parameters

Parameter	Input/Output	Description
original_model_file	Input	Definition file (.pb) of the dequantized TensorFlow model. A string.
quantize_model_file	Input	Definition file (.pb) of the quantized TensorFlow model. A string.
outputs	Input	String list of the output node. A string.
sampler_config_file	Input	Performance sampler configuration file provided by users. For details about this file and the configuration example, see Performance Sampler Configuration File. A string.
save_dir	Input	Path for saving the model after performance unquantization. A string.
strategy	Input	Policy for searching for quantization configurations that meet performance requirements. A string or a Python instance (PerfStrategyBase). Default: BatchRollBack

Returns

None

Example

import amct_tensorflow as amct
def main():
    args_check(args)
    model_file = args.model_file_name
    quant_model_file = args.quant_model_file_name
    outputs = args.save_outputs
    save_dir = args.path_dir
    sampler_config_file = args.sampler_config_file
 
    amct.perf_based_auto_calibration(
        model_file, quant_model_file, outputs[:], sampler_config_file, save_dir)

Flush files:

A .pb model file that can be used for accuracy simulation in the TensorFlow environment or inference on the AI processor.
A quantization factor record file, a quantization configuration file, a performance comparison file, and an automatic unquantization history file.

Parent topic: PTQ APIs