Zhonghui You

Probably because some filters that did not work during the observe process were cut out by mistake. The easiest solution is to increase OBSERVE_TIMES in prune/universal.py line 13. The definition...

Removing a single filter in the group convolution in PyTorch will cause misalignment, but it's possible to remove the entire group by using the Group Pruning method proposed in our...

If you remove a whole group of filters at a time, the number can be kept divisible. The troublesome thing is to maintain the number of channels of the input...

如果最后一层是卷积层，应该也是可以已被 Conv2dObserver 处理的。

In the current version of the code, the module we implement is GBN(Gated BatchNorm). You can implement the GC(Gated Convolution) module according to the GBN code. We tried GC on...

Download from here -> [cub.zip](https://github.com/youzhonghui/gate-decorator-pruning/files/4276951/cub.zip) This notebook shows the code and running results of the CUB experimental in the paper. But because it relies on the old version gbn implementation,...