fix bug in 3d group convolution #9860
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@stephenyan1231 could you also add appropriate test cases to cover your fix? |
stephenyan1231
force-pushed
the
export-D9008807
branch
from
e954c81 to
6ddd3bf
Compare
Summary: Pull Request resolved: pytorch#9860 For 3D group convolution, in the case of CUDNN 7 and NCHWD order, filter dim is (M, C/group_, k_h, h_w, k_d). According to CUDA doc (https://docs.nvidia.com/deeplearning/sdk/cudnn-developer-guide/index.html#grouped-convolutions), the existing implementation is incorrect, and will crash the 3d video model training with group convolution. In the implementation, `filter.dims(1)` is already `C/group_`. So don't need to divide it by `group_` again. Differential Revision: D9008807 fbshipit-source-id: 5c2ebe12a0db875d593e0457136aaab6ecf27660
stephenyan1231
force-pushed
the
export-D9008807
branch
from
6ddd3bf to
4d64a67
Compare
xiaomengy
approved these changes
Jul 31, 2018
goodlux
pushed a commit
to goodlux/pytorch
that referenced
this pull request
Aug 15, 2018
Summary: Pull Request resolved: pytorch#9860 For 3D group convolution, in the case of CUDNN 7 and NCHWD order, filter dim is (M, C/group_, k_h, h_w, k_d). According to CUDA doc (https://docs.nvidia.com/deeplearning/sdk/cudnn-developer-guide/index.html#grouped-convolutions), the existing implementation is incorrect, and will crash the 3d video model training with group convolution. In the implementation, `filter.dims(1)` is already `C/group_`. So don't need to divide it by `group_` again. Reviewed By: BIT-silence Differential Revision: D9008807 fbshipit-source-id: 2f0d6eb47f4e16d7417a7e3baeba709e3254154f
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
For 3D group convolution, in the case of CUDNN 7 and NCHWD order, filter dim is (M, C/group_, k_h, h_w, k_d).
According to CUDA doc (https://docs.nvidia.com/deeplearning/sdk/cudnn-developer-guide/index.html#grouped-convolutions), the existing implementation is incorrect, and will crash the 3d video model training with group convolution.
In the implementation,
filter.dims(1)is alreadyC/group_. So don't need to divide it bygroup_again.Differential Revision: D9008807