Skip to content
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 Convolution2D.backward_gpu so that it does not use cuda.linalg #341

Merged
merged 2 commits into from
Aug 25, 2015

Conversation

delta2323
Copy link
Member

When CuDNN is not used, Convolution2D.backward_gpu uses cuda.linalg, which is no longer available after CuPy is introduced. This PR removes dependency on cuda.linalg.dot and cuda.linalg.add_dot and uses cuda.cupy.dot instead.

@beam2d beam2d added the cat:bug Bug report or fix. label Aug 24, 2015
@beam2d beam2d added this to the v1.3.0 milestone Aug 24, 2015
@beam2d beam2d self-assigned this Aug 24, 2015
@beam2d
Copy link
Member

beam2d commented Aug 24, 2015

We should fix the test code to pass the im2col/col2im path even in cuDNN-enabled environments.

@beam2d
Copy link
Member

beam2d commented Aug 25, 2015

I fixed tests to check im2col version even if cuDNN is enabled.

@delta2323
Copy link
Member Author

As for @beam2d's fix, it looks good for me.

@beam2d
Copy link
Member

beam2d commented Aug 25, 2015

OK, then I'll merge.

beam2d added a commit that referenced this pull request Aug 25, 2015
Fix Convolution2D.backward_gpu so that it does not use cuda.linalg
@beam2d beam2d merged commit 8f34325 into master Aug 25, 2015
@beam2d beam2d deleted the fix_convolution2d branch August 25, 2015 08:33
@beam2d
Copy link
Member

beam2d commented Aug 25, 2015

Thank you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
cat:bug Bug report or fix.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants