Skip to content

Conversation

@xiaomengy
Copy link
Contributor

Summary: Update Im2Col related to make preparation for group conv in NHWC order.

Differential Revision: D9285344

@xiaomengy xiaomengy requested a review from houseroad August 11, 2018 19:56
@xiaomengy
Copy link
Contributor Author

@pytorchbot retest this please

2 similar comments
@xiaomengy
Copy link
Contributor Author

@pytorchbot retest this please

@xiaomengy
Copy link
Contributor Author

@pytorchbot retest this please

Copy link
Member

@houseroad houseroad left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, no need to do the loop on the group, which is great.

This comment was marked as off-topic.

This comment was marked as off-topic.

@jspark1105
Copy link
Contributor

jspark1105 commented Aug 14, 2018

@BIT-silence thanks for the PR. This should help what I'm working on a lot. One comment. Using batched gemm for group convolution in NHCW definitely increases parallelism a lot, but at the same time it increases working set size (we're generating G times bigger col_buffer) so I think we should check how it affects the performance. CC @ajtulloch

This comment was marked as off-topic.

This comment was marked as off-topic.

This comment was marked as off-topic.

@fmassa fmassa added the caffe2 label Aug 14, 2018
…r. (pytorch#10439)

Summary:
Pull Request resolved: pytorch#10439

Update Im2Col related to make preparation for group conv in NHWC order.

Reviewed By: houseroad

Differential Revision: D9285344

fbshipit-source-id: eb4c581daf3e95050de6d8b1bdaf567164e7e1cd
@xiaomengy xiaomengy deleted the export-D9285344 branch August 17, 2018 18:21
@ezyang ezyang added the merged label Jun 26, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants