How to perform Global Pruning with constant pruning ratio per layer

[sudaksh14]

Hi, I am working on an application where I need to record all the indexes which are pruned in each step of iterative pruning. Now when I use the global pruning approach, it does not prune each layer equally and hence I can't create a dense model from it due to shape mismatch of layers after pruning. This leaves me with local pruning, but in this case the indexes are duplicated in each round because the resulting model has layers with smaller shapes of the weights. How can I keep pruning globally but with same pruning rate for each layer (Eg: 20% neurons of each layer). Thanks in advanced!