create_quant_retrain_model

Applicability

Product	Supported
Atlas A3 training series products/Atlas A3 inference series products	x
Atlas A2 training products/Atlas A2 inference products	x
Atlas 200I/500 A2 inference product	√
Atlas inference series products	√
Atlas training products	√

Description

Performs quantization on a graph based on the configuration file (config_file), inserts activation and weight fake-quantization layers, and saves the modified network to a new model file.

Prototype

create_quant_retrain_model(model_file, weights_file, config_file, modified_model_file, modified_weights_file, scale_offset_record_file)

Parameters

Parameter	Input/Output	Description
model_file	Input	Definition file (.prototxt) of the Caffe model. A string.
weights_file	Input	Weight file (.caffemodel) of the trained Caffe model. A string.
config_file	Input	Quantization configuration file. A string.
modified_model_file	Input	Name of the resultant Caffe model definition file (.prototxt) for storing the inserted QAT layers. A string.
modified_weights_file	Input	Name of the resultant Caffe model weight file (.caffemodel) for storing the inserted QAT layers. A string.
scale_offset_record_file	Input	File for storing quantization factors. A string.

Returns

None

Example

from amct_caffe import amct
model_file = 'resnet50_train.prototxt'
weights_file = 'ResNet-50-model.caffemodel'
modified_model_file = './tmp/modified_model.prototxt'
modified_weights_file = './tmp/modified_model.caffemodel'
config_json_file = './config.json'
scale_offset_record_file = './record.txt'
amct.create_quant_retrain_model( model_file, weights_file, config_json_file,  modified_model_file, modified_weights_file, scale_offset_record_file)

Flush files:

modified_model_file: definition file of the modified model, with QAT layers inserted into the original model.
modified_weights_file: weight file of the modified model, with QAT layers inserted into the original model.

Parent topic: QAT APIs