Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
ignore_index defined but not used
See original GitHub issueignore_indexis defined but not used.
I assume the purpose is to ignore the padding from the text. If that’s the case, then instead of using:
F.cross_entropy(rearrange(logits, 'b n c -> b c n'), labels)
We have to define a nn.Module Cross_entropy
self.loss = nn.CrossEntropyLoss(reduction='none', ignore_index=self.ignore_index)
and then
self.loss(rearrange(logits, 'b n c -> b c n'), labels)
Finally the default value of ignore_index should be 0. Since thats the pad_id of the default tokenizer.
Issue Analytics
- State:
- Created 3 years ago
- Reactions:1
- Comments:5 (3 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
I just realized. We would ignore also the VQVAE codebook with id of 0. Should we just change the padding in the tokenizer to a negative value?
Can probably close this one, right? Sorry, the issues are super scattered currently.