What is -eq in tlt yolo_v4 prune

I found this explanation:
-eq, --equalization_criterion : Criteria to equalize the stats of inputs to an element-wise op layer or depth-wise convolutional layer. This parameter is useful for ResNets and MoblieNets. Options are arithmetic_mean ,:code:geometric_mean, union , and intersection . The default value is union .
but I still don’t understand why it is necessary and what each option does.


See Frequently Asked Questions — Transfer Learning Toolkit 3.0 documentation,

equalization_criterion is to choose method to merge weights from different branches of element wise or depth wise layers, default is union

Is there any webpage that explains the different methods?

Sorry for late reply.
Please see PRUNING NEURAL NETWORKS THAT INCLUDE ELEMENT-WISE OPERATIONS and Pruning Models with NVIDIA Transfer Learning Toolkit | NVIDIA Developer Blog