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

[LLM Inference] add --use_fake_parameter option for ptq fake scales and fix compute error of total_max_length #8955

Merged
merged 6 commits into from
Aug 19, 2024

Conversation

yuanlehome
Copy link
Collaborator

@yuanlehome yuanlehome commented Aug 17, 2024

PR types

Others

PR changes

Others

Description

  • add --use_fake_parameter option for ptq fake scales
  • 修复src_length与max_length的问题,现在导出静态图模型时无须指定它们。
  • 修复llama3/qwen2等模型输出尾部多余的“#”符号
  • 更新tune_cublaslt_gemm示例,增加llama3.1/qwen2 tune示例

Copy link

paddle-bot bot commented Aug 17, 2024

Thanks for your contribution!

Copy link

codecov bot commented Aug 17, 2024

Codecov Report

Attention: Patch coverage is 0% with 73 lines in your changes missing coverage. Please review.

Project coverage is 54.82%. Comparing base (e0d2809) to head (0cb99af).
Report is 225 commits behind head on develop.

Files with missing lines Patch % Lines
...dlenlp/experimental/transformers/llama/modeling.py 0.00% 38 Missing ⚠️
paddlenlp/experimental/transformers/utils.py 0.00% 34 Missing ⚠️
paddlenlp/utils/llm_utils.py 0.00% 1 Missing ⚠️
Additional details and impacted files
@@             Coverage Diff             @@
##           develop    #8955      +/-   ##
===========================================
+ Coverage    54.79%   54.82%   +0.02%     
===========================================
  Files          636      636              
  Lines        99876    99970      +94     
===========================================
+ Hits         54732    54807      +75     
- Misses       45144    45163      +19     

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

@yuanlehome yuanlehome changed the title [LLM Inference] update some code [LLM Inference] add --use_fake_parameter option for ptq fake scales and fix compute error of total_max_length Aug 17, 2024
@@ -124,7 +124,7 @@ class PredictorArgument:

@property
def total_max_length(self):
return self.src_length + self.max_length
return 8192 # Maximum sequence length.
Copy link
Contributor

Choose a reason for hiding this comment

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

why?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

类似模型的超参,之前理解有问题,因此改正过来

Copy link
Contributor

@DesmonDay DesmonDay 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 71b3be3 into PaddlePaddle:develop Aug 19, 2024
9 of 12 checks passed
Mangodadada pushed a commit to Mangodadada/PaddleNLP that referenced this pull request Sep 10, 2024
…nd fix compute error of total_max_length (PaddlePaddle#8955)

* update some code

* update

* update

* update

* update tune_cublaslt_gemm demo

* fix step in tune_cublaslt_gemm
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.

3 participants