[Application] Distill and finetune LLM

[DrownFish19]

Before submitting

• Lint code. If there are lint issues, please format the code first.

# Install and register `pre-commit` in the project folder
pip install pre-commit && pre-commit install

# Process previous code files separately
pre-commit run --file XXXX.py

• Add test cases into tests folder. If there are codecov issues, please add tests cases first.

PR types

New features

PR changes

Others

Description

Add an application to distill and finetune LLM.

• step1: prepare data.
• step2: distill new data from reasoning models.
• stpe3: fintune the LLM.
• step4: eval the LLM with GSM8k, AIME2024 and MATH-500.

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 50.20%. Comparing base (8452021) to head (a395f97).
Report is 5 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop   #10061      +/-   ##
===========================================
- Coverage    50.21%   50.20%   -0.02%     
===========================================
  Files          757      757              
  Lines       122403   122448      +45     
===========================================
+ Hits         61467    61476       +9     
- Misses       60936    60972      +36     

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

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ZHUI]

https://github.com/PaddlePaddle/PaddleNLP/tree/develop/llm/benchmark 推理这里搞了一版评估。