Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Minor fix to reenable nvtx sequence numbers for the forward methods of custom (Python) autograd functions#13876

Closed
mcarilli wants to merge 2 commits into
pytorch:masterfrom
mcarilli:nvtx_custom_fwd
Closed

Minor fix to reenable nvtx sequence numbers for the forward methods of custom (Python) autograd functions#13876
mcarilli wants to merge 2 commits into
pytorch:masterfrom
mcarilli:nvtx_custom_fwd

Conversation

@mcarilli
Copy link
Copy Markdown
Collaborator

@mcarilli mcarilli commented Nov 13, 2018

Some of our arch people (@mkolod, Aditya Agrawal, @kevinstephano) notified me that the sequence number annotations weren't showing up for forward methods of custom autograd functions, which was breaking their nvprof dump parsing. Two one-line fixes in the appropriate code paths.

Copy link
Copy Markdown
Contributor

@ezyang ezyang left a comment

Choose a reason for hiding this comment

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

No chance we could get a test, could we?

Copy link
Copy Markdown
Contributor

@facebook-github-bot facebook-github-bot left a comment

Choose a reason for hiding this comment

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

@ezyang is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

@mcarilli
Copy link
Copy Markdown
Collaborator Author

@ezyang I'm not sure how. A test would have to do something like run nvprof on a standalone file to completion, spit out an nvvp dump, then open it (separately) and parse the sqlite database. Any suggestions?

@ezyang
Copy link
Copy Markdown
Contributor

ezyang commented Nov 13, 2018

That would probably be the most likely way to work. Maybe not worth the effort. But I promise you we will accidentally break it without any tests at some point in the future, if someone else touches this code ;)

laurentdupin pushed a commit to laurentdupin/pytorch that referenced this pull request Apr 24, 2026
…f custom (Python) autograd functions (pytorch#13876)

Summary:
Some of our arch people (mkolod, Aditya Agrawal, kevinstephano) notified me that the sequence number annotations weren't showing up for forward methods of custom autograd functions, which was breaking their nvprof dump parsing.  Two one-line fixes in the appropriate code paths.
Pull Request resolved: pytorch#13876

Differential Revision: D13042381

Pulled By: ezyang

fbshipit-source-id: a114118f5c07ad4ba482e7a4892d08805b23c65b
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants