Benefit and usage of torch.nn.utils.prune

[ani0075 (Aniruddha Saha)]

Check my answer here.

How to improve inference time of pruned model using torch.nn.utils.prune?

As @ypsoh mentioned, the torch.sparse module is still in beta. So you won’t get any speed up for a model pruned in an unstructured way unless you make use of sparse matrix operations. Hopefully torch.sparse will be part of a stable release soon.