Fix rank_zero_only rank not set in ddp-spawn based strategies
#19030
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
awaelchli
requested review from
carmocca,
justusschock,
Borda and
williamFalcon
as code owners
awaelchli
added
bug
github-actions
bot
added
fabric
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #19030 +/- ##
==========================================
- Coverage 84% 54% -30%
==========================================
Files 443 438 -5
Lines 36154 36060 -94
==========================================
- Hits 30260 19461 -10799
- Misses 5894 16599 +10705 |
for more information, see https://pre-commit.ci
awaelchli
added
the
fun
Borda
approved these changes
Nov 20, 2023
Borda
pushed a commit
that referenced
this pull request
Dec 19, 2023
(cherry picked from commit f652e6c)
lantiga
pushed a commit
that referenced
this pull request
Dec 20, 2023
(cherry picked from commit f652e6c)
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Outdated suggestions cannot be applied.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.
What does this PR do?
This PR temporarily fixes the issue that the DDP strategies who call
set_world_ranksonly set therank_zero_only.rankattribute for the fabric/pytorch utilities, however NOT for thelightning_utilitiespackage. This leads to logs appearing on rank > 0 on spawn-based strategies.The decision to outsource these utilities into the separate package made all of this very brittle. Now we need to maintain 2 globals instead of one. With this fix, I will open an issue requesting that these utilities get moved back into the Lightning package so we only need to maintain one global variable (rank_zero_only.rank).
Minimal repro (just observe info and warn outputs duplicated):
Discovered while debugging CI flakiness. This fix should automatically resolve the timeouts for some of the ddp-spawn tests like
test_result_reduce_ddpetc.📚 Documentation preview 📚: https://pytorch-lightning--19030.org.readthedocs.build/en/19030/
cc @Borda @tchaton @carmocca @justusschock @awaelchli