Skip to content

[DAPO] Fix data preprocess script #172

Merged
lynnliu030 merged 16 commits intoNovaSky-AI:mainfrom
lynnliu030:main
Aug 20, 2025
Merged

[DAPO] Fix data preprocess script #172
lynnliu030 merged 16 commits intoNovaSky-AI:mainfrom
lynnliu030:main

Conversation

@lynnliu030
Copy link
Member

No description provided.

@lynnliu030 lynnliu030 requested a review from SumanthRH August 20, 2025 07:52
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

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

Code Review

This pull request introduces a new AIME environment for math execution tasks, along with data preparation and training scripts. The changes are generally well-structured. My feedback focuses on improving robustness, efficiency, and correctness. Key points include:

  • Using proper error handling instead of assertions in the AIMEEnv.
  • Correcting a type hint in the aime.utils module.
  • Improving the efficiency of the data preprocessing script by using polars for I/O directly.
  • Making the data download script idempotent.
  • Enhancing the main run script for better error handling and to include the data preprocessing step.

Copy link
Member

Choose a reason for hiding this comment

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

can you rename this to data_preprocess_aime?

Or rather, we could just have a separate aime folder in examples

Copy link
Member Author

Choose a reason for hiding this comment

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

done, I just make it data_preprocess_dapo_aime.py

@lynnliu030 lynnliu030 merged commit a4ae67d into NovaSky-AI:main Aug 20, 2025
3 checks passed
dzorlu referenced this pull request in fleet-ai/SkyRL Feb 4, 2026
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.

2 participants