Quantization Configuration Parameters

**Table 1** retrain_enable
Description	QAT enable
Type	bool
Value	true or false
Command-Line Options	true: on false: off
Recommended	true
Required/Optional	This function is optional.

**Table 2** retrain_data_config
Description	Activation quantization configuration
Type	dict
Value	-
Command-Line Options	Includes the following parameters: batch_num: batch number used for quantization. clip_max: upper bound of clipping-based quantization, defaulted to be empty. clip_min: lower bound of clipping-based quantization, defaulted to be empty. fixed_min: whether to fix the minimum value of clipping-based quantization to 0, defaulted to be empty. dst_type: Currently, only INT8 quantization bit width is supported. The default value is INT8.
Recommended Value	-
Required/Optional	This function is optional.

**Table 3** retrain_weight_config
Description	Weight quantization configuration
Specification	dict
Value	-
Command-Line Options	Includes the following parameters: algo: quantization algorithm select, defaulted to arq_retrain. Whether to use different quantization factors for each channel. dst_type: Currently, only INT8 quantization bit width is supported. The default value is INT8.
Recommended Value	-
Required/Optional	This function is optional.

**Table 4** weights_retrain_algo parameters
Description	Quantization algorithm
Type	string
Value	-
Command-Line Options	ulq_quantize: ULQ clipping-based quantization algorithm arq_retrain: ARQ quantization algorithm
Recommended Value	arq_retrain
Required/Optional	Optional

**Table 5** channel_wise
Description	Whether to use different quantization factors for each channel
Specification	bool
Value	true or false
Command-Line Options	true: Channels are separately quantized using different quantization factors. false: All channels are quantized altogether using the same quantization factors.
Recommended	true
Required/Optional	This function is optional.

**Table 6** batch_num
Description	Batch count used for quantization.
"Type"	int
The options are as follows:	Greater than 0
Command-line options	Defaults to 1. You are advised to keep the calibration dataset size within 50 images. Calculate batch_num based on batch_size as follows: batch_num x batch_size = Calibration dataset size batch_size indicates the number of images per batch.
Recommended Value	1
Required/Optional	This function is optional.

**Table 7** fixed_min
Description	Fixed lower bound switch for the activation quantization algorithm
Type	bool
Value	true or false
Parameter Description	true: fixes the lower bound of the activation quantization algorithm at 0. false: does not fix the lower bound of the activation quantization algorithm.
Recommended Value	Do not include this parameter.
Required/Optional	This function is optional.

**Table 8** clip_max
Description	Upper bound for the activation quantization algorithm
Type	float
Value	clip_max>0 Find the maximum max based on the distribution of the activation values at different layers. The recommended value range is [0.3 x max, 1.7 x max].
Command-Line Options	If this parameter is included, the upper bound of the clipping-based activation quantization algorithm is fixed. If this parameter is not included, the upper bound is learned using the IFMR algorithm.
Recommended Value	Do not include this parameter.
Required/Optional	This function is optional.

**Table 9** clip_min
Description	Lower bound for the activation quantization algorithm
Type	float
Value	clip_min<0 Find the minimum min based on the distribution of the activation values at different layers. The recommended value range is [0.3 x min, 1.7 x min].
Command-Line Options	If this parameter is included, the lower bound of the clipping-based activation quantization algorithm is fixed. If this parameter is not included, the lower bound is learned using the IFMR algorithm.
Recommended Value	Do not include this parameter.
Required/Optional	This function is optional.

**Table 10** dst_type
Description	Quantization bit width select
Type	string
Value	Currently, only INT8 (default) is supported.
Command-Line Options	Selects the quantization bit width during quantization.
Recommended Value	-
Required/Optional	This function is optional.

Parent topic: Single-operator mode