Skip to content

Raise an error if target is out-of-bounds in ClassNLLCriterion #5299

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
apaszke merged 1 commit into from
Feb 19, 2018
Merged

Raise an error if target is out-of-bounds in ClassNLLCriterion #5299

apaszke merged 1 commit into from
Feb 19, 2018

Conversation

@elanmart
Copy link
Contributor

@elanmart elanmart commented Feb 19, 2018

This PR is supposed to fix #5296

New behaviour:

>>> import torch
>>> from torch.autograd import Variable
>>> import torch.nn.functional as F

>>> p = Variable(torch.randn(3, 3), requires_grad=True)
>>> y = Variable(torch.LongTensor([99, 100, 101]))
>>> l = F.cross_entropy(p, y, reduce=False)

RuntimeError: Target 100 out of bounds at /.../pytorch/aten/src/THNN/generic/ClassNLLCriterion.c:56

I'm not sure why *_no_reduce cuda kernels did not take n_classes as argument, was it meant to be that way?

@onnxbot-worker-3 onnxbot-worker-3 mentioned this pull request Feb 19, 2018
Closed
@elanmart
Copy link
Contributor Author

elanmart commented Feb 19, 2018

@pytorchbot retest this please

@apaszke apaszke merged commit 6279367 into pytorch:master Feb 19, 2018
@apaszke
Copy link
Contributor

apaszke commented Feb 19, 2018

Thanks Marcin!

@elanmart elanmart deleted the fix-nll-criterion branch February 19, 2018 16:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@apaszke apaszke apaszke approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

open source

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

reduce=False --> No error for invalid labels

3 participants

@elanmart @apaszke @ezyang