Skip to main content

Model Compression

Path: Artificial Intelligence --- Code Development and Training --- Model Compression

alt text

Create a Compression Task​

Input Model: Click the selection button and choose the model you need from the pop-up model list. After selection, the relevant information of this model will be automatically filled in. It should be noted that only models with configured compression information can be selected here. You can configure the compression - related information in the model version details section. For more details, see the Model Version Details chapter.

Compression Method: Depending on the model configuration, it can support compression methods such as quantization, pruning, and knowledge distillation.

Compression Parameters: For different compression methods and configurations, you can fill in the corresponding compression parameters.

Dataset: Select the dataset required for model compression. It is used to evaluate and guide the compression algorithm, aiming to maintain model performance while reducing model size and computational cost.