--compress_weight_conf

Applicability

Description

Sets the directory of the node list configuration file to be compressed, including the file name.

See Also

None

Arguments

Argument: Directory (including the file name) of the node list configuration file to be compressed. Nodes are mainly Conv and FC operators.

Format: The directory (including the file name) can contain letters, digits, underscores (_), hyphens (-), periods (.), and Chinese characters.

Restrictions: The operator list in the weight compression configuration file is output by the AMCT (in the file that records non-uniformly quantized layers in the directory of the non-uniform quantization result, for example, module/results/calibration_results/module_nuq_layer_record.txt). The file is a list of node names, which are separated by semicolons (;).

Suggestions and Benefits

None

Examples

For example, the content of the compress_weight_nodes.cfg file is: conv1;fc1;conv2_2/x1;fc2;conv5_32/x2;fc6

Run the following command to upload the file to the server where the AOE is deployed. /home/Davinci is used as an example.

--compress_weight_conf=/home/Davinci/compress_weight_nodes.cfg

Dependencies and Restrictions

When using this option, ensure that the model in use is the deployment model generated after the non-uniform quantization operation is performed by AMCT.