Hi,
Is masking the padding as shown here (under Trick 3) still relevant in PyTorch 0.4.1, or does the ignore_index
parameter of nn.CrossEntropyLoss
takes care of that for me?
Thanks!
Hi,
Is masking the padding as shown here (under Trick 3) still relevant in PyTorch 0.4.1, or does the ignore_index
parameter of nn.CrossEntropyLoss
takes care of that for me?
Thanks!