itay hubara

Hi, I am trying to use the code for SVHN normalization (GCN+LCN) and save the results. I receive that the validation dataset (i.e ds:validSet():inputs():input()) is the same size as the...

wave2vec2 with feat_extract_norm == "group" normalizes over channels not token which causes issue when padding

6

### System Info Any GPU/CPU system ### Who can help? _No response_ ### Information While there is support for padding, I believe it affects the accuracy - (even when running...

reduce padding overhead when using buckets

1

## 🚀 Feature Request The current get_buckets function ensures that you have the same number of samples in each bucket. This can result in many unnecessary padding. The feature would...