create_quant_config
Description
Finds all quantization-capable layers, including InnerProduct, Convolution (using a 1-dilated 4 x 4 convolution filter), Deconvolution (using a 1-dilated 4 x 4 convolution filter with group = 1), and AVE Pooling, automatically generates a quantization configuration file, and writes the quantization configuration information of the quantization-capable layers to the configuration file.
Restrictions
Due to data type conversion, the quantization factors in the generated quantization configuration file might be different from those in the simplified configuration file. However, the accuracy is not affected.
Prototype
create_quant_config(config_file, model_file, weights_file, skip_layers=None, batch_num=1, activation_offset=True, config_defination=None)
Parameters
Parameter |
Input/Return |
Description |
Restrictions |
|---|---|---|---|
config_file |
Input |
Directory of the quantization configuration file, including the file name. The existing file (if available) in the directory will be overwritten upon this API call. |
A string. |
model_file |
Input |
Caffe model definition file (.prototxt). |
A string. |
weights_file |
Input |
Weight file of the already-trained Caffe model (.caffemodel). |
A string. |
skip_layers |
Input |
Quantization-capable layers to skip. |
Default: None A list of strings. |
batch_num |
Input |
Number of batches taken to generate the quantization factors. |
An int. Value range: any integer larger than 0. Default: 1 batch_num cannot be too large. The product of batch_num and batch_size equals to the number of images used during quantization. Too many images consume too much memory. |
activation_offset |
Input |
Whether to quantize data with offset. |
Default: true A bool. |
config_defination |
Input |
Whether to create a simplified quantization configuration file quant.cfg from the calibration_config.proto file in /amct_caffe/proto/calibration_config.proto in the AMCT installation path. For details about the available options in the calibration_config.proto file and the generated simplified quantization configuration file quant.cfg, see Simplified Quantization Configuration File. |
Default: None A string. If set to None, a configuration file is generated based on the residual arguments (skip_layers, batch_num, and activation_offset). Otherwise, a configuration file in JSON format is generated based on this argument. |
Returns
None
Outputs
Outputs a quantization configuration file in JSON format. (When quantization is performed again, the existing quantization configuration file output by this API call will be overwritten.)
For example, if only layer_name1 and layer_name2 in a graph support quantization, the quantization configuration file generated by the create_quant_config(config_file, model_file, weights_file, skip_layers=None, batch_num=2, activation_offset=True, config_defination=None) call is as follows.
{
"version":1,
"batch_num":2,
"activation_offset":true,
"do_fusion":true,
"skip_fusion_layers":[],
"layer_name1":{
"quant_enable":true,
"activation_quant_params":{
"max_percentile":0.999999,
"min_percentile":0.999999,
"search_range":[
0.7,
1.3
],
"search_step":0.01
},
"weight_quant_params":{
"wts_algo":"arq_quantize",
"channel_wise":true
}
},
"layer_name2":{
"quant_enable":true,
"activation_quant_params":{
"max_percentile":0.999999,
"min_percentile":0.999999,
"search_range":[
0.7,
1.3
],
"search_step":0.01
},
"weight_quant_params":{
"wts_algo":"arq_quantize",
"channel_wise":false
}
}
}
Example
from amct_caffe import create_quant_config
# Create a quantization configuration file from arguments.
create_quant_config(config_file="./configs/config.json",
model_file="./pretrained_model/model.prototxt",
weights_file="./pretrained_model/model.caffemodel",
skip_layers=None,
batch_num=1,
activation_offset=True)
# Generate a quantization configuration file from the simplified configuration file.
create_quant_config(config_file="./configs/config.json",
model_file="./pretrained_model/model.prototxt",
weights_file="./pretrained_model/model.caffemodel",
config_defination="./configs/quant.cfg")
