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

add llama & qwen dpo #8474

Merged
merged 6 commits into from
Jun 11, 2024
Merged

add llama & qwen dpo #8474

merged 6 commits into from
Jun 11, 2024

Conversation

lugimzzz
Copy link
Contributor

PR types

New features

PR changes

APIs

Description

新增dpo

Copy link

paddle-bot bot commented May 21, 2024

Thanks for your contribution!

Copy link

codecov bot commented May 21, 2024

Codecov Report

Attention: Patch coverage is 8.24176% with 501 lines in your changes are missing coverage. Please review.

Project coverage is 54.02%. Comparing base (b36b6a0) to head (affd27d).
Report is 5 commits behind head on develop.

Current head affd27d differs from pull request most recent head 2854cf1

Please upload reports for the commit 2854cf1 to get more accurate results.

Files Patch % Lines
paddlenlp/trl/dpo_trainer.py 7.85% 352 Missing ⚠️
paddlenlp/trl/trl_data.py 3.66% 105 Missing ⚠️
paddlenlp/datasets/zero_padding_dataset.py 17.85% 23 Missing ⚠️
paddlenlp/trl/trl_utils.py 4.54% 21 Missing ⚠️
Additional details and impacted files
@@             Coverage Diff             @@
##           develop    #8474      +/-   ##
===========================================
- Coverage    54.29%   54.02%   -0.27%     
===========================================
  Files          617      621       +4     
  Lines        96339    96878     +539     
===========================================
+ Hits         52310    52343      +33     
- Misses       44029    44535     +506     

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

llm/utils.py Outdated
@@ -211,13 +211,6 @@ def prediction_step(
# keepdim in order to maintain the same shape as logits
if isinstance(logits, (list, tuple)):
logits = logits[0]
# all gather logits when enabling tensor_parallel_output
Copy link
Collaborator

Choose a reason for hiding this comment

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

这里删除tensor parallel output的支持的原因是什么?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

删错了,需要恢复

Copy link
Collaborator

@wawltor wawltor left a comment

Choose a reason for hiding this comment

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

LGTM

@wawltor wawltor merged commit 909be01 into PaddlePaddle:develop Jun 11, 2024
8 of 10 checks passed
@lugimzzz lugimzzz deleted the dpo branch June 27, 2024 07:25
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.

None yet

2 participants