ColaCheng

Results 1 comments of ColaCheng

A trick solution for too large value (which causes `nan` output) ```python is_bad = torch.logical_or(torch.abs(feat) > 10, torch.isnan(feat)) feat[is_bad] = 0 ```