ruifanxu
Results
1
comments of
ruifanxu
The data pass into the function has padded already, so those sentences less than max length will be padded with 0. Before constructing the mask, `pad = torch.eq(seqs, self.padding_idx)` will...