Better bounds checks in ctcloss#16269
Conversation
|
cc @t-vi |
t-vi
left a comment
There was a problem hiding this comment.
Thanks for sending this!
The code changes themselves look good to me, I think it would be good to add a few more tests.
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
Tests are failing now |
|
I dont know why |
|
I wouldn't worry too much about the failure for 31fdb04 - the rocm CI builds seem to be more prone to un-related build errors (and unlike the circle ci, I'm not sure they automatically retry). |
|
seems like failure on the circle ci |
|
So now it all seems good for me, the build failure is indeed unrelated. Thanks @ashermancinelli ! |
|
@t-vi Yep! Let me know if there's anything else I need to do to get it merged. |
facebook-github-bot
left a comment
There was a problem hiding this comment.
@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Summary: Adds better bounds checks for target lengths in CTC loss, checks for integral types for target and prediction lengths, and adds tests for each, according to #15946 Pull Request resolved: pytorch/pytorch#16269 Differential Revision: D13847567 Pulled By: ezyang fbshipit-source-id: 5d7a975565e02baf78fe388813a1d1ef56dfb212
Summary: Adds better bounds checks for target lengths in CTC loss, checks for integral types for target and prediction lengths, and adds tests for each, according to pytorch#15946 Pull Request resolved: pytorch#16269 Differential Revision: D13847567 Pulled By: ezyang fbshipit-source-id: 5d7a975565e02baf78fe388813a1d1ef56dfb212
Adds better bounds checks for target lengths in CTC loss, checks for integral types for target and prediction lengths, and adds tests for each, according to #15946