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

Fix sharding reshard bug and support reshard in different partition ways #8837

Conversation

sneaxiy
Copy link
Collaborator

@sneaxiy sneaxiy commented Jul 30, 2024

PR types

Bug fixes

PR changes

Others

Description

  • Fix bugs of sharding reshard when tensor name is not kept after spawning another processes.
  • Make sharding reshard support automatic reshard when the tensors are partitioned in different ways.

Copy link

paddle-bot bot commented Jul 30, 2024

Thanks for your contribution!

Copy link

codecov bot commented Jul 30, 2024

Codecov Report

Attention: Patch coverage is 9.45946% with 67 lines in your changes missing coverage. Please review.

Project coverage is 55.21%. Comparing base (2a57101) to head (0c1e6ab).
Report is 34 commits behind head on incubate/paddlenlp-fleety.

Files with missing lines Patch % Lines
paddlenlp/trainer/utils/reshard/sharding_v2.py 11.11% 40 Missing ⚠️
paddlenlp/trainer/utils/sharding_io.py 6.25% 15 Missing ⚠️
paddlenlp/trainer/trainer.py 7.69% 12 Missing ⚠️
Additional details and impacted files
@@                      Coverage Diff                      @@
##           incubate/paddlenlp-fleety    #8837      +/-   ##
=============================================================
- Coverage                      55.24%   55.21%   -0.04%     
=============================================================
  Files                            620      620              
  Lines                          96574    96636      +62     
=============================================================
+ Hits                           53351    53355       +4     
- Misses                         43223    43281      +58     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@sneaxiy sneaxiy requested a review from ForFishes July 30, 2024 15:00
@sneaxiy sneaxiy merged commit 99d3d93 into PaddlePaddle:incubate/paddlenlp-fleety Jul 31, 2024
5 of 7 checks passed
@sneaxiy sneaxiy deleted the sharding_reshard_enhance branch July 31, 2024 06:05
sneaxiy added a commit to sneaxiy/PaddleNLP that referenced this pull request Aug 2, 2024
…ays (PaddlePaddle#8837)

* Fix sharding reshard bug and support reshard in different partition ways

* fix some bugs

* refine codes

* revert some useless codes

* add log
wawltor pushed a commit that referenced this pull request Aug 6, 2024
* Fix sharding reshard bug and support reshard in different partition ways (#8837)

* Fix sharding reshard bug and support reshard in different partition ways

* fix some bugs

* refine codes

* revert some useless codes

* add log

* Fix TP sync (#8846)

* fix sharding reshard compatiblity for wrong master weight name (#8851)

* change print to logger.info

* remove useless f-string
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant