let `_get_default_process_group_backend_for_device` support more hardware platforms #21057

taozhiwei · 2025-08-12T07:55:41Z

This change is to enable _get_default_process_group_backend_for_device to support more hardware platforms and no longer hard code for cuda

📚 Documentation preview 📚: https://pytorch-lightning--21057.org.readthedocs.build/en/21057/

Borda

looks good, just please add a test with mock if needed :)

codecov · 2025-08-12T15:22:20Z

Codecov Report

❌ Patch coverage is 75.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 87%. Comparing base (5a2b678) to head (5f43a0c).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #21057   +/-   ##
=======================================
- Coverage      87%      87%   -0%     
=======================================
  Files         268      268           
  Lines       23367    23370    +3     
=======================================
+ Hits        20379    20381    +2     
- Misses       2988     2989    +1

taozhiwei · 2025-08-13T07:27:48Z

looks good, just please add a test with mock if needed :)

I have added the test, and it seems that the last CI failure was caused by the running environment, could you please help trigger it again? Thank you

tests/tests_fabric/utilities/test_distributed.py

…l _get_default_process_group_backend_for_device Signed-off-by: taozhiwei <[email protected]>

for more information, see https://pre-commit.ci

src/lightning/fabric/utilities/distributed.py

tests/tests_fabric/utilities/test_distributed.py

Co-authored-by: Nicki Skafte Detlefsen <[email protected]>

taozhiwei · 2025-08-14T02:44:48Z

It seems this CI failure is not related to this change. Can you help trigger another CI? Thank you
@SkafteNicki @bhimrazy

taozhiwei · 2025-08-14T07:25:34Z

https://github.com/Lightning-AI/pytorch-lightning/pull/21057/checks?check_run_id=48055959528
https://github.com/Lightning-AI/pytorch-lightning/pull/21057/checks?check_run_id=48055963335
The latest trigger has failed again, it still seems to be an environmental issue?

SkafteNicki · 2025-08-14T11:14:01Z

@taozhiwei yes it seems unrelated to this issue. When we get it fixed the PR will be merged.

taozhiwei · 2025-08-15T07:30:50Z

yes it seems unrelated to this issue. When we get it fixed the PR will be merged.

Thank you very much !

…ore hardware platforms (#21057)" This reverts commit 119a640.

Borda · 2025-08-18T10:12:35Z

needs to be reverted and added again as this breaks DDP fork...
@taozhiwei could you pls try to add it again 🦩

…1057 (#21092) * debug failing tests for Fabric with `ddp_fork` on PT 2.8 * Revert "let `_get_default_process_group_backend_for_device` support more hardware platforms (#21057)" This reverts commit 119a640. --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…ware platforms (#21057) * support more hardware platforms and no longer hard code cuda when call _get_default_process_group_backend_for_device * Apply suggestions from code review --------- Signed-off-by: taozhiwei <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Nicki Skafte Detlefsen <[email protected]> Co-authored-by: Jirka Borovec <[email protected]> (cherry picked from commit 119a640)

…1057 (#21092) * debug failing tests for Fabric with `ddp_fork` on PT 2.8 * Revert "let `_get_default_process_group_backend_for_device` support more hardware platforms (#21057)" This reverts commit 119a640. --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> (cherry picked from commit 9ca360b)

…ware platforms (#21057) * support more hardware platforms and no longer hard code cuda when call _get_default_process_group_backend_for_device * Apply suggestions from code review --------- Signed-off-by: taozhiwei <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Nicki Skafte Detlefsen <[email protected]> Co-authored-by: Jirka Borovec <[email protected]> (cherry picked from commit 119a640)

…1057 (#21092) * debug failing tests for Fabric with `ddp_fork` on PT 2.8 * Revert "let `_get_default_process_group_backend_for_device` support more hardware platforms (#21057)" This reverts commit 119a640. --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> (cherry picked from commit 9ca360b)

* let `_get_default_process_group_backend_for_device` support more hardware platforms (#21057) * support more hardware platforms and no longer hard code cuda when call _get_default_process_group_backend_for_device * Apply suggestions from code review --------- * try it * chlog --------- Signed-off-by: taozhiwei <[email protected]> Co-authored-by: taozhiwei <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Nicki Skafte Detlefsen <[email protected]>

* let `_get_default_process_group_backend_for_device` support more hardware platforms (#21057) * support more hardware platforms and no longer hard code cuda when call _get_default_process_group_backend_for_device * Apply suggestions from code review --------- * try it * chlog --------- Signed-off-by: taozhiwei <[email protected]> Co-authored-by: taozhiwei <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Nicki Skafte Detlefsen <[email protected]> (cherry picked from commit 3c81316)

taozhiwei requested review from lantiga, Borda, tchaton, justusschock and ethanwharris as code owners August 12, 2025 07:55

github-actions bot added the fabric lightning.fabric.Fabric label Aug 12, 2025

Borda changed the title ~~let _get_default_process_group_backend_for_device support more hardware platforms~~ let _get_default_process_group_backend_for_device support more hardware platforms Aug 12, 2025

Borda reviewed Aug 12, 2025

View reviewed changes

taozhiwei force-pushed the myfeature branch from a56756f to 3ed2485 Compare August 13, 2025 07:25

taozhiwei force-pushed the myfeature branch from dccc839 to 116bd74 Compare August 13, 2025 07:43

bhimrazy reviewed Aug 13, 2025

View reviewed changes

tests/tests_fabric/utilities/test_distributed.py Outdated Show resolved Hide resolved

bhimrazy approved these changes Aug 13, 2025

View reviewed changes

support more hardware platforms and no longer hard code cuda when cal…

cd22015

…l _get_default_process_group_backend_for_device Signed-off-by: taozhiwei <[email protected]>

taozhiwei force-pushed the myfeature branch from 6937e72 to cd22015 Compare August 13, 2025 08:27

pre-commit-ci bot and others added 2 commits August 13, 2025 08:27

[pre-commit.ci] auto fixes from pre-commit.com hooks

64ac379

for more information, see https://pre-commit.ci

Merge branch 'master' into myfeature

20e6fe6

SkafteNicki reviewed Aug 13, 2025

View reviewed changes

src/lightning/fabric/utilities/distributed.py Show resolved Hide resolved

tests/tests_fabric/utilities/test_distributed.py Show resolved Hide resolved

taozhiwei changed the title ~~let _get_default_process_group_backend_for_device support more hardware platforms~~ let _get_default_process_group_backend_for_device support more hardware platforms Aug 13, 2025

Borda changed the title ~~let _get_default_process_group_backend_for_device support more hardware platforms~~ let _get_default_process_group_backend_for_device support more hardware platforms Aug 13, 2025

Apply suggestions from code review

84d2e47

Co-authored-by: Nicki Skafte Detlefsen <[email protected]>

SkafteNicki approved these changes Aug 13, 2025

View reviewed changes

Merge branch 'master' into myfeature

5f43a0c

Borda approved these changes Aug 15, 2025

View reviewed changes

Borda merged commit 119a640 into Lightning-AI:master Aug 15, 2025
117 of 120 checks passed

Borda added a commit that referenced this pull request Aug 18, 2025

Revert "let _get_default_process_group_backend_for_device support m…

169f69c

…ore hardware platforms (#21057)" This reverts commit 119a640.

Borda mentioned this pull request Aug 18, 2025

debug failing tests for Fabric with ddp_fork on PT 2.8 -> revert #21057 #21092

Merged

7 tasks

Borda mentioned this pull request Aug 18, 2025

debug failing tests for Fabric with ddp_fork on PT 2.8 #21093

Merged

7 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

let `_get_default_process_group_backend_for_device` support more hardware platforms #21057

let `_get_default_process_group_backend_for_device` support more hardware platforms #21057

Uh oh!

taozhiwei commented Aug 12, 2025 •

edited by github-actions bot

Loading

Uh oh!

Borda left a comment

Uh oh!

codecov bot commented Aug 12, 2025 •

edited

Loading

Uh oh!

taozhiwei commented Aug 13, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

taozhiwei commented Aug 14, 2025

Uh oh!

taozhiwei commented Aug 14, 2025 •

edited

Loading

Uh oh!

SkafteNicki commented Aug 14, 2025

Uh oh!

taozhiwei commented Aug 15, 2025

Uh oh!

Uh oh!

Borda commented Aug 18, 2025 •

edited

Loading

Uh oh!

Uh oh!

let _get_default_process_group_backend_for_device support more hardware platforms #21057

let _get_default_process_group_backend_for_device support more hardware platforms #21057

Uh oh!

Conversation

taozhiwei commented Aug 12, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Borda left a comment

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Aug 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

taozhiwei commented Aug 13, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

taozhiwei commented Aug 14, 2025

Uh oh!

taozhiwei commented Aug 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

SkafteNicki commented Aug 14, 2025

Uh oh!

taozhiwei commented Aug 15, 2025

Uh oh!

Uh oh!

Borda commented Aug 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

let `_get_default_process_group_backend_for_device` support more hardware platforms #21057

let `_get_default_process_group_backend_for_device` support more hardware platforms #21057

taozhiwei commented Aug 12, 2025 •

edited by github-actions bot

Loading

codecov bot commented Aug 12, 2025 •

edited

Loading

taozhiwei commented Aug 14, 2025 •

edited

Loading

Borda commented Aug 18, 2025 •

edited

Loading