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

[DCU] fix DCU w8a8c8 GEMM shape #9115

Merged
merged 1 commit into from
Sep 11, 2024

Conversation

YanhuiDua
Copy link
Contributor

@YanhuiDua YanhuiDua commented Sep 10, 2024

PR types

Bug fixes

PR changes

Others

Description

fix DCU GEMM shape when quant_type == "a8w8c8"

Copy link

paddle-bot bot commented Sep 10, 2024

Thanks for your contribution!

Copy link

codecov bot commented Sep 10, 2024

Codecov Report

Attention: Patch coverage is 0% with 7 lines in your changes missing coverage. Please review.

Project coverage is 53.33%. Comparing base (2f31866) to head (81c30d8).
Report is 6 commits behind head on develop.

Files with missing lines Patch % Lines
...erimental/transformers/fused_transformer_layers.py 0.00% 4 Missing ⚠️
...dlenlp/experimental/transformers/llama/modeling.py 0.00% 1 Missing ⚠️
...enlp/experimental/transformers/mixtral/modeling.py 0.00% 1 Missing ⚠️
...dlenlp/experimental/transformers/qwen2/modeling.py 0.00% 1 Missing ⚠️
Additional details and impacted files
@@             Coverage Diff             @@
##           develop    #9115      +/-   ##
===========================================
- Coverage    53.34%   53.33%   -0.01%     
===========================================
  Files          652      652              
  Lines       105401   105404       +3     
===========================================
  Hits         56222    56222              
- Misses       49179    49182       +3     

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

@@ -631,6 +631,7 @@ def __init__(self, config: LlamaConfig):
)

else:
print("self.quant_type: ", self.quant_type)
Copy link
Collaborator

Choose a reason for hiding this comment

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

注释删掉

Copy link
Contributor Author

Choose a reason for hiding this comment

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

done

@@ -372,7 +372,7 @@ def __init__(self, config: Qwen2Config):
use_neox_rotary_style=self.use_neox,
cachekv_int8_type=config.cachekv_int8_type,
rank_id=config.tensor_parallel_rank,
trans_qkvw=(False if paddle.is_compiled_with_rocm() and self.quant_type == "a8w8" else True),
trans_qkvw=(False if paddle.is_compiled_with_rocm() and "a8w8" in self.quant_type else True),
Copy link
Contributor Author

Choose a reason for hiding this comment

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

这个就是为了修复w8a8c8时的gemm shape ~

@YanhuiDua YanhuiDua closed this Sep 11, 2024
@YanhuiDua YanhuiDua reopened this Sep 11, 2024
@CLAassistant
Copy link

CLAassistant commented Sep 11, 2024

CLA assistant check
All committers have signed the CLA.

@qingqing01 qingqing01 merged commit 73a3db9 into PaddlePaddle:develop Sep 11, 2024
6 of 12 checks passed
@YanhuiDua YanhuiDua deleted the fix_dcu_gemm branch September 11, 2024 13:18
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.

4 participants