Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Support PyTorch Lightning's FSDP optimizer states saving and loading #17819

Merged
merged 21 commits into from
Jul 7, 2023
Merged

Support PyTorch Lightning's FSDP optimizer states saving and loading #17819

merged 21 commits into from
Jul 7, 2023
+215 −12

Conversation

leng-yue
Copy link
Contributor

@leng-yue leng-yue commented Jun 13, 2023 •
edited by awaelchli
Loading

What does this PR do?

Similar to the FSDP optimizer states fix #17747, this pull request makes saving and restoring FSDP optimizer states on lightning possible.

Fixes #17798
Fixes #17702
Fixes #16815

Before submitting
  • Was this discussed/agreed via a GitHub issue? (not for typos and docs)
  • Did you read the contributor guideline, Pull Request section?
  • Did you make sure your PR does only one thing, instead of bundling different changes together?
  • Did you make sure to update the documentation with your changes? (if necessary)
  • Did you write any new necessary tests? (not for typos and docs)
  • Did you verify new and existing tests pass locally with your changes?
  • Did you list all the breaking changes introduced by this pull request?
  • Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

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
  • Is this pull request ready for review? (if not, please submit in draft mode)
  • Check that all items from Before submitting are resolved
  • Make sure the title is self-explanatory and the description concisely explains the PR
  • Add labels and milestones (and optionally projects) to the PR so it can be classified

@github-actions github-actions bot added the pl Generic label for PyTorch Lightning package label Jun 13, 2023
@leng-yue leng-yue marked this pull request as ready for review June 13, 2023 02:53
@github-actions github-actions bot added the fabric lightning.fabric.Fabric label Jun 13, 2023
@awaelchli awaelchli self-assigned this Jun 20, 2023
@leng-yue
Copy link
Contributor Author

Is there anything I can do?

@awaelchli awaelchli added this to the 2.1 milestone Jun 25, 2023
awaelchli
awaelchli reviewed Jun 25, 2023
View reviewed changes
Copy link
Member

@awaelchli awaelchli left a comment

Choose a reason for hiding this comment

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

Thanks for the help @leng-yue. Great work! Apologies for the late replies (on vacation).

tests/tests_pytorch/strategies/test_fsdp.py Show resolved Hide resolved
tests/tests_pytorch/strategies/test_fsdp.py Show resolved Hide resolved
src/lightning/fabric/strategies/fsdp.py Outdated Show resolved Hide resolved
@awaelchli awaelchli added strategy: fsdp Fully Sharded Data Parallel checkpointing Related to checkpointing labels Jul 3, 2023
@mergify mergify bot removed the has conflicts label Jul 3, 2023
@mergify mergify bot added the has conflicts label Jul 3, 2023
@mergify mergify bot removed the has conflicts label Jul 3, 2023
@mergify mergify bot added the has conflicts label Jul 4, 2023
@mergify mergify bot removed the has conflicts label Jul 4, 2023
@awaelchli awaelchli added the community This PR is from the community label Jul 4, 2023
carmocca
carmocca approved these changes Jul 5, 2023
View reviewed changes
Copy link
Contributor

@carmocca carmocca left a comment

Choose a reason for hiding this comment

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

Great work!

src/lightning/pytorch/strategies/fsdp.py Outdated Show resolved Hide resolved
src/lightning/pytorch/strategies/fsdp.py Show resolved Hide resolved
tests/tests_pytorch/strategies/test_fsdp.py Outdated Show resolved Hide resolved
tests/tests_pytorch/strategies/test_fsdp.py Outdated Show resolved Hide resolved
tests/tests_pytorch/strategies/test_fsdp.py Show resolved Hide resolved
src/lightning/pytorch/strategies/fsdp.py Outdated Show resolved Hide resolved
tests/tests_pytorch/strategies/test_fsdp.py Outdated Show resolved Hide resolved
@mergify mergify bot added the ready PRs ready to be merged label Jul 5, 2023
@awaelchli awaelchli enabled auto-merge (squash) July 7, 2023 17:29
@awaelchli awaelchli merged commit 734a325 into Lightning-AI:master Jul 7, 2023
101 checks passed
@awaelchli
Copy link
Member

Thank you @leng-yue for your patience. Great progress on this awesome feature!

@awaelchli awaelchli mentioned this pull request Aug 4, 2023
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@carmocca carmocca carmocca approved these changes

@awaelchli awaelchli awaelchli approved these changes

@justusschock justusschock Awaiting requested review from justusschock justusschock is a code owner

@williamFalcon williamFalcon Awaiting requested review from williamFalcon

Assignees

@awaelchli awaelchli

Labels
checkpointing Related to checkpointing community This PR is from the community fabric lightning.fabric.Fabric pl Generic label for PyTorch Lightning package ready PRs ready to be merged strategy: fsdp Fully Sharded Data Parallel
Projects
None yet
Milestone
2.1
4 participants
@leng-yue @awaelchli @carmocca @Borda