admm-pruning
admm-pruning copied to clipboard
Prune DNN using Alternating Direction Method of Multipliers (ADMM)
I have made a pytorch version. It also allows filter and column pruning
I have a question: how to determine the prune ratio of each layer? by try-and-error?
我写了一个简单的pytorch实现,但是结果很奇怪,最近您是否有空给我一点指导?
Is a PyTorch implementation going to be provided?
When I change your code to pytorch, I found the loss doesn't convergence, and the conditions W-Z is also doesn't convergence.
I read the codes in it and find that they just set the connection weights as zeros. And then use the whole matrix to calculate the output. I assume that...