Support saving sharded checkpoints with FSDP in Trainer #18364
Merged
Conversation
This file contains 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
for more information, see https://pre-commit.ci
…to feature/dist-checkpoint-fsdp
awaelchli
force-pushed
the
feature/dist-checkpoint-fsdp-save
branch
from
9a1e324 to
049299d
Compare
for more information, see https://pre-commit.ci
…ing-AI/lightning into feature/dist-checkpoint-fsdp-save
for more information, see https://pre-commit.ci
awaelchli
added
feature
for more information, see https://pre-commit.ci
awaelchli
commented
Aug 22, 2023
awaelchli
requested review from
carmocca,
justusschock,
williamFalcon,
tchaton and
Borda
as code owners
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #18364 +/- ##
=========================================
- Coverage 84% 62% -23%
=========================================
Files 427 422 -5
Lines 33158 33077 -81
=========================================
- Hits 28011 20433 -7578
- Misses 5147 12644 +7497 |
Borda
approved these changes
Aug 22, 2023
carmocca
approved these changes
Aug 22, 2023
for more information, see https://pre-commit.ci
…e' into feature/dist-checkpoint-fsdp-save
awaelchli
commented
Aug 23, 2023
7 tasks
This was referenced Aug 23, 2023
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Outdated suggestions cannot be applied.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?
Part 1 of supporting sharded/distributed checkpoints. This PR implements the saving part. The loading part turns out to be a bit more complex and needs more debugging and testing (#18358).
The code is very similar to the one in Fabric's FSDP strategy with minor changes to support the checkpoint dict structure that the checkpoint connector provides.
The "sharded" state dict format can be enabled via:
The default remains "full" for now since loading is not yet implemented and to stay backward compatible. The documentation will receive an update once the loading part has also landed.
cc @Borda @carmocca @justusschock @awaelchli