machine learning - Writing Custom Python Layer With Learnable Parameters in Caffe -

i know this example supposed illustrate how add trainable parameters in python layer using add_blob() method.

however, still unable understand how can used set dimensions of blob based on user defined parameters.

there better example here on how write python layer here. here, layer not contain trainable parameters.

please explain how write custom python layer trainable parameters.

when add parameters blob using add_blob(), can reshape added blob, either in setup() method (right when add it), or in layer's reshape() method.