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

Fix initialization of optimizers in DDP Strategy #11952

Merged
merged 31 commits into from
Jun 1, 2022
Merged

Fix initialization of optimizers in DDP Strategy #11952

merged 31 commits into from
Jun 1, 2022

Conversation

ananthsub
Copy link
Contributor

@ananthsub ananthsub commented Feb 17, 2022 •
edited by rohitgr7
Loading

What does this PR do?

Redo of #11886
also fixes: #12317

Does your PR introduce any breaking changes? If yes, please list them.

Before submitting

  • Was this discussed/approved 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 internal minor changes/refactorings)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

  • 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

Did you have fun?

Make sure you had fun coding 🙃

@ananthsub ananthsub added bug Something isn't working strategy: ddp DistributedDataParallel strategy: ddp spawn strategy: fairscale sharded (removed) Sharded Data Parallel labels Feb 17, 2022
@ananthsub ananthsub changed the title Create optimizers in DDP strategy after moving model to device Move model to device before creating optimizers in DDP Feb 17, 2022
@DuYicong515 DuYicong515 mentioned this pull request Feb 18, 2022
Closed
10 tasks
awaelchli
awaelchli approved these changes Feb 19, 2022
View reviewed changes
pytorch_lightning/strategies/sharded.py Outdated Show resolved Hide resolved
@mergify mergify bot added ready PRs ready to be merged has conflicts labels Feb 19, 2022
@ananthsub ananthsub force-pushed the fix/ddp-delay-optimizer-creation branch from c90cad9 to 46fd630 Compare March 5, 2022 06:35
@mergify mergify bot removed the has conflicts label Mar 5, 2022
@ananthsub ananthsub changed the title Move model to device before creating optimizers in DDP Optimize initialization of optimizers in strategies Mar 5, 2022
ananthsub
ananthsub commented Mar 5, 2022
View reviewed changes
pytorch_lightning/strategies/ddp.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/fully_sharded.py Show resolved Hide resolved
pytorch_lightning/strategies/ddp_spawn.py Outdated Show resolved Hide resolved
tchaton
tchaton reviewed Mar 5, 2022
View reviewed changes
pytorch_lightning/strategies/ddp.py Show resolved Hide resolved
pytorch_lightning/strategies/ddp.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/ddp.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/ddp_spawn.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/ddp_spawn.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/fully_sharded.py Show resolved Hide resolved
awaelchli
awaelchli reviewed Mar 6, 2022
View reviewed changes
tests/strategies/test_ddp_strategy.py Show resolved Hide resolved
pytorch_lightning/strategies/ddp.py Show resolved Hide resolved
pytorch_lightning/strategies/ddp.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/ddp_spawn.py Outdated Show resolved Hide resolved
pytorch_lightning/strategies/tpu_spawn.py Show resolved Hide resolved
pytorch_lightning/strategies/ddp.py Outdated Show resolved Hide resolved
@ananthsub ananthsub force-pushed the fix/ddp-delay-optimizer-creation branch from cb4cb5b to a456b1b Compare March 8, 2022 06:43
@ananthsub ananthsub changed the title Optimize initialization of optimizers in strategies Fix initialization of optimizers in DDP Strategy Mar 8, 2022
@ananthsub ananthsub mentioned this pull request Mar 8, 2022
Merged
12 tasks
@mergify mergify bot added has conflicts and removed ready PRs ready to be merged labels May 11, 2022
@rohitgr7 rohitgr7 enabled auto-merge (squash) May 25, 2022 12:09
@mergify mergify bot added ready PRs ready to be merged and removed has conflicts ready PRs ready to be merged labels May 25, 2022
@austinmw
Copy link

Any updates on this PR?

@mergify mergify bot added has conflicts and removed ready PRs ready to be merged labels Jun 1, 2022
@mergify mergify bot added ready PRs ready to be merged and removed has conflicts ready PRs ready to be merged labels Jun 1, 2022
@rohitgr7 rohitgr7 merged commit c1f0502 into Lightning-AI:master Jun 1, 2022
@carmocca
Copy link
Contributor

carmocca commented Jun 1, 2022

@austinmw Merged! Will be included in the 1.6.4 release

carmocca added a commit that referenced this pull request Jun 1, 2022
@ananthsub @rohitgr7 @carmocca
Fix initialization of optimizers in DDP Strategy (#11952)
4b8b10c 
Co-authored-by: Rohit Gupta <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
carmocca added a commit that referenced this pull request Jun 1, 2022
@ananthsub @rohitgr7 @carmocca
Fix initialization of optimizers in DDP Strategy (#11952)
acc6339 
Co-authored-by: Rohit Gupta <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
lexierule pushed a commit that referenced this pull request Jun 1, 2022
@ananthsub @rohitgr7
@carmocca @lexierule
Fix initialization of optimizers in DDP Strategy (#11952)
a5f82f5 
Co-authored-by: Rohit Gupta <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>
@rohitgr7 rohitgr7 mentioned this pull request Jul 1, 2022
Merged
12 tasks
@awaelchli awaelchli mentioned this pull request Aug 12, 2022
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@tchaton tchaton tchaton left review comments

@awaelchli awaelchli awaelchli approved these changes

@carmocca carmocca carmocca approved these changes

@Borda Borda Borda approved these changes

@SeanNaren SeanNaren Awaiting requested review from SeanNaren

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

@kaushikb11 kaushikb11 Awaiting requested review from kaushikb11

@williamFalcon williamFalcon Awaiting requested review from williamFalcon

@rohitgr7 rohitgr7 Awaiting requested review from rohitgr7

Assignees

@rohitgr7 rohitgr7

Labels
bug Something isn't working ready PRs ready to be merged strategy: ddp DistributedDataParallel strategy: fairscale sharded (removed) Sharded Data Parallel
Projects
None yet
Milestone
pl:1.6.x
Development

Successfully merging this pull request may close these issues.

Can't use estimated_stepping_batches in configure_optimizers with DDP
8 participants
@ananthsub @rohitgr7 @dawnos @austinmw @carmocca @awaelchli @Borda @tchaton