-
Notifications
You must be signed in to change notification settings - Fork 21.4k
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.
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
Fix cuDNN error message when it's Conv2d #45729
Conversation
The weird error message before this PR.
Note these The |
cc @ptrblck |
Codecov Report
@@ Coverage Diff @@
## master #45729 +/- ##
=======================================
Coverage 68.60% 68.60%
=======================================
Files 410 410
Lines 52670 52670
=======================================
Hits 36132 36132
Misses 16538 16538 Continue to review full report at Codecov.
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@ngimel has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
Summary: Originally introduced in pytorch#45023. When I was doing test in the original PR, it was a Conv3d, so this problem was not discovered. Arrays in `ConvolutionParams` have a fixed length of 3 or 5. This is because `max_dim` is set as a constexpr of 3, regardless of Conv2d or Conv3d. The current code will make some error message be weird. See below in the comments. https://github.com/pytorch/pytorch/blob/9201c37d020007979e144693d86c8e8599e2fd8f/aten/src/ATen/native/cudnn/Conv.cpp#L212-L226 Pull Request resolved: pytorch#45729 Reviewed By: mruberry Differential Revision: D24081542 Pulled By: ngimel fbshipit-source-id: 141f8946f4d0db63a723131775731272abeaa6ab
Summary: Originally introduced in #45023. When I was doing test in the original PR, it was a Conv3d, so this problem was not discovered. Arrays in `ConvolutionParams` have a fixed length of 3 or 5. This is because `max_dim` is set as a constexpr of 3, regardless of Conv2d or Conv3d. The current code will make some error message be weird. See below in the comments. https://github.com/pytorch/pytorch/blob/9201c37d020007979e144693d86c8e8599e2fd8f/aten/src/ATen/native/cudnn/Conv.cpp#L212-L226 Pull Request resolved: #45729 Reviewed By: mruberry Differential Revision: D24081542 Pulled By: ngimel fbshipit-source-id: 141f8946f4d0db63a723131775731272abeaa6ab
Originally introduced in #45023. When I was doing test in the original PR, it was a Conv3d, so this problem was not discovered.
Arrays in
ConvolutionParams
have a fixed length of 3 or 5. This is becausemax_dim
is set as a constexpr of 3, regardless of Conv2d or Conv3d. The current code will make some error message be weird. See below in the comments.pytorch/aten/src/ATen/native/cudnn/Conv.cpp
Lines 212 to 226 in 9201c37