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.

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

Distributed sampling parity between Lite and PyTorch #16101

Merged
merged 14 commits into from
Dec 19, 2022

Conversation

awaelchli
Copy link
Contributor

@awaelchli awaelchli commented Dec 17, 2022

What does this PR do?

Restores parity between pure PyTorch dataloader and Lite dataloader when using DistributedSampler (DDP).
This PR avoids the DistributedSamplerWrapper for the two standard samplers, RandomSampler and SequentialSampler. The DistributedSamplerWrapper should only be used for custom samplers. A note was added to the documentation of this wrapper.

We also ensure that .set_epoch() gets called automatically for the user.

Fixes #14636

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

The order of samples when shuffling is enabled. But this should be considered a bugfix.

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 update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
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?

I made sure I had fun coding 🙃

cc @Borda @carmocca @justusschock @awaelchli

@awaelchli awaelchli added feature Is an improvement or enhancement fabric lightning.fabric.Fabric fun Staff contributions outside working hours - to differentiate from the "community" label labels Dec 17, 2022
@awaelchli awaelchli added bug Something isn't working and removed feature Is an improvement or enhancement labels Dec 17, 2022
@awaelchli awaelchli added this to the v1.8.x milestone Dec 17, 2022
@awaelchli awaelchli marked this pull request as ready for review December 17, 2022 13:43
@github-actions
Copy link
Contributor

github-actions bot commented Dec 17, 2022

⚡ Required checks status: All passing 🟢

Groups summary

🟢 pytorch_lightning: Tests workflow
Check ID Status
pl-cpu (macOS-11, pytorch, 3.8, 1.11) success
pl-cpu (macOS-11, pytorch, 3.9, 1.12) success
pl-cpu (macOS-11, pytorch, 3.10, 1.13) success
pl-cpu (macOS-11, pytorch, 3.8, 1.10, oldest) success
pl-cpu (ubuntu-20.04, pytorch, 3.8, 1.10) success
pl-cpu (ubuntu-20.04, pytorch, 3.9, 1.11) success
pl-cpu (ubuntu-20.04, pytorch, 3.10, 1.12) success
pl-cpu (ubuntu-20.04, pytorch, 3.10, 1.13) success
pl-cpu (ubuntu-20.04, pytorch, 3.7, 1.10, oldest) success
pl-cpu (windows-2022, pytorch, 3.9, 1.11) success
pl-cpu (windows-2022, pytorch, 3.10, 1.12) success
pl-cpu (windows-2022, pytorch, 3.10, 1.13) success
pl-cpu (windows-2022, pytorch, 3.7, 1.10, oldest) success
pl-cpu (slow, macOS-11, pytorch, 3.7, 1.11) success
pl-cpu (slow, ubuntu-20.04, pytorch, 3.7, 1.11) success
pl-cpu (slow, windows-2022, pytorch, 3.7, 1.11) success
pl-cpu (macOS-11, lightning, 3.8, 1.13) success
pl-cpu (ubuntu-20.04, lightning, 3.8, 1.13) success
pl-cpu (windows-2022, lightning, 3.8, 1.13) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.

🟢 pytorch_lightning: Azure GPU
Check ID Status
pytorch-lightning (GPUs) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.

🟢 pytorch_lightning: Azure HPU
Check ID Status
pytorch-lightning (HPUs) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.

🟢 pytorch_lightning: Azure IPU
Check ID Status
pytorch-lightning (IPUs) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.

🟢 lightning_lite: CPU workflow
Check ID Status
lite-cpu (macOS-11, lite, 3.8, 1.11) success
lite-cpu (macOS-11, lite, 3.9, 1.12) success
lite-cpu (macOS-11, lite, 3.10, 1.13) success
lite-cpu (macOS-11, lite, 3.7, 1.10, oldest) success
lite-cpu (ubuntu-20.04, lite, 3.8, 1.10) success
lite-cpu (ubuntu-20.04, lite, 3.9, 1.11) success
lite-cpu (ubuntu-20.04, lite, 3.10, 1.12) success
lite-cpu (ubuntu-20.04, lite, 3.10, 1.13) success
lite-cpu (ubuntu-20.04, lite, 3.7, 1.10, oldest) success
lite-cpu (windows-2022, lite, 3.9, 1.11) success
lite-cpu (windows-2022, lite, 3.10, 1.12) success
lite-cpu (windows-2022, lite, 3.10, 1.13) success
lite-cpu (windows-2022, lite, 3.7, 1.10, oldest) success
lite-cpu (macOS-11, lightning, 3.8, 1.13) success
lite-cpu (ubuntu-20.04, lightning, 3.8, 1.13) success
lite-cpu (windows-2022, lightning, 3.8, 1.13) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py, tests/tests_lite/test_lite.py, tests/tests_lite/test_wrappers.py.

🟢 lightning_lite: Azure GPU
Check ID Status
lightning-lite (GPUs) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py, tests/tests_lite/test_lite.py, tests/tests_lite/test_wrappers.py.

🟢 mypy
Check ID Status
mypy success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.

🟢 install
Check ID Status
install-pkg (ubuntu-22.04, app, 3.7) success
install-pkg (ubuntu-22.04, app, 3.10) success
install-pkg (ubuntu-22.04, lite, 3.7) success
install-pkg (ubuntu-22.04, lite, 3.10) success
install-pkg (ubuntu-22.04, pytorch, 3.7) success
install-pkg (ubuntu-22.04, pytorch, 3.10) success
install-pkg (ubuntu-22.04, lightning, 3.7) success
install-pkg (ubuntu-22.04, lightning, 3.10) success
install-pkg (ubuntu-22.04, notset, 3.7) success
install-pkg (ubuntu-22.04, notset, 3.10) success
install-pkg (macOS-12, app, 3.7) success
install-pkg (macOS-12, app, 3.10) success
install-pkg (macOS-12, lite, 3.7) success
install-pkg (macOS-12, lite, 3.10) success
install-pkg (macOS-12, pytorch, 3.7) success
install-pkg (macOS-12, pytorch, 3.10) success
install-pkg (macOS-12, lightning, 3.7) success
install-pkg (macOS-12, lightning, 3.10) success
install-pkg (macOS-12, notset, 3.7) success
install-pkg (macOS-12, notset, 3.10) success
install-pkg (windows-2022, app, 3.7) success
install-pkg (windows-2022, app, 3.10) success
install-pkg (windows-2022, lite, 3.7) success
install-pkg (windows-2022, lite, 3.10) success
install-pkg (windows-2022, pytorch, 3.7) success
install-pkg (windows-2022, pytorch, 3.10) success
install-pkg (windows-2022, lightning, 3.7) success
install-pkg (windows-2022, lightning, 3.10) success
install-pkg (windows-2022, notset, 3.7) success
install-pkg (windows-2022, notset, 3.10) success

These checks are required after the changes to src/lightning_lite/lite.py, src/lightning_lite/utilities/distributed.py, src/lightning_lite/wrappers.py.


Thank you for your contribution! 💜

Note
This comment is automatically generated and updates for 60 minutes every 180 seconds. If you have any other questions, contact carmocca for help.

@awaelchli awaelchli modified the milestones: v1.8.x, v1.9 Dec 17, 2022
@awaelchli awaelchli added feature Is an improvement or enhancement and removed bug Something isn't working labels Dec 17, 2022
@mergify mergify bot added the ready PRs ready to be merged label Dec 19, 2022
Copy link
Contributor

@tchaton tchaton left a comment

Choose a reason for hiding this comment

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

LGTM !

src/lightning_lite/wrappers.py Outdated Show resolved Hide resolved
src/lightning_lite/wrappers.py Outdated Show resolved Hide resolved
src/lightning_lite/wrappers.py Outdated Show resolved Hide resolved
tests/tests_lite/test_lite.py Outdated Show resolved Hide resolved
@awaelchli awaelchli enabled auto-merge (squash) December 19, 2022 21:20
@awaelchli awaelchli merged commit 7bc55c3 into master Dec 19, 2022
@awaelchli awaelchli deleted the bugfix/ddp-sampler-parity branch December 19, 2022 21:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
fabric lightning.fabric.Fabric feature Is an improvement or enhancement fun Staff contributions outside working hours - to differentiate from the "community" label ready PRs ready to be merged
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Introduce a mechanism to set the epoch on the sampler in LightningLite
5 participants