Flexible and easy to use HSDP setting
What does this PR do?
Fixes #19502
It allows users to provide device_mesh as a tuple to the FSDPStrategy.
The test case for hybrid fsdp strategy would be broken, because _init_sharding_strategy is moved out of the __init__ function.
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
📚 Documentation preview 📚: https://pytorch-lightning--19504.org.readthedocs.build/en/19504/
Codecov Report
Attention: Patch coverage is 66.66667% with 6 lines in your changes missing coverage. Please review.
Project coverage is 59%. Comparing base (
1a6786d) to head (949d36f). Report is 97 commits behind head on master.
:exclamation: There is a different number of reports uploaded between BASE (1a6786d) and HEAD (949d36f). Click for more details.
HEAD has 188 uploads less than BASE
Flag BASE (1a6786d) HEAD (949d36f) python3.10 23 15 cpu 91 43 lightning 55 31 pytest 71 26 examples 9 0 app 9 0 lightning_fabric 19 8 python3.11 34 16 python3.8 18 12 pytorch2.0 12 11 pytest-full 24 21 pytorch_lightning 10 8 pytorch2.1 6 5 pytorch2.3 3 2 lightning_app 2 0
Additional details and impacted files
@@ Coverage Diff @@
## master #19504 +/- ##
=========================================
- Coverage 84% 59% -25%
=========================================
Files 426 421 -5
Lines 35264 35182 -82
=========================================
- Hits 29553 20772 -8781
- Misses 5711 14410 +8699
Any updates on this?
Hi @awaelchli , I just fixed a merge conflict with the master and pushed the change. Could you take another look? Let me know if you have any concerns.
All checks have passed. But I don't have access to merge the PR.