Improve RichProgreasBar speed performance #21321
Open
+58
−5
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Outdated suggestions cannot be applied.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.
What does this PR do?
Attempt to fix #13937
Benchmark Results are as below
Key changes:
CustomProgress.refreshnow only calls its superclass method when running in interactive mode. This change is critical for performance since in most cases, we only need tonotifythe refresh thread rather than performing a full refresh.threading.Eventfor custom_RefreshThreadUsing
Eventwould introduce unnecessary locking overhead. Since acquiring and releasing locks are relatively expensive here, this approach didn’t yield any speed improvement.Simply enabling
auto_refreshwould revert Addrefresh_rateto RichProgressBar #10497 and bring back pdb - based debugging with RichProgressBar doesn't show input #10607, which causes pdb unusable when usingRichProgressBar.TBH, I’m not entirely sure if this is the best solution, but I haven’t found another way to improve performance without undoing prior fixes.
Given that we’re now defaulting to rich-based progress bars and model summaries when rich is installed, I think this could still be quite a beneficial improvement — happy to hear thoughts or alternative ideas.
Before submitting
PR review
Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:
Reviewer checklist
📚 Documentation preview 📚: https://pytorch-lightning--21321.org.readthedocs.build/en/21321/