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 Linear Weight Initialization in the PaddlePaddle Implementation #1413

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

GuoxiaWang
Copy link

Description

fix linear weight initilize for paddle framework

Type of change

  • Documentation change (change only to the documentation, either a fix or a new content)
  • Bug fix (non-breaking change which fixes an issue)
  • New feature (non-breaking change which adds functionality)
  • Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • Infra/Build change
  • Code refactoring

Changes

In the PaddlePaddle implementation of TE, the weights of Linear use a column-major order, whereas paddle.nn.Linear in the PaddlePaddle framework uses a row-major order. This inconsistency in behavior can result in different initialization values between te.Linear and paddle.nn.Linear. This PR addresses and corrects this initialization logic.

Checklist:

  • I have read and followed the contributing guidelines
  • The functionality is complete
  • I have commented my code, particularly in hard-to-understand areas
  • I have made corresponding changes to the documentation
  • My changes generate no new warnings
  • I have added tests that prove my fix is effective or that my feature works
  • New and existing unit tests pass locally with my changes

@GuoxiaWang GuoxiaWang force-pushed the fix_paddle_init branch 2 times, most recently from 26abdd3 to 4c9419a Compare January 21, 2025 10:10
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