search

xingyaoww / code-act

Official Repo for ICML 2024 paper "Executable Code Actions Elicit Better LLM Agents" by Xingyao Wang, Yangyi Chen, Lifan Yuan, Yizhe Zhang, Yunzhu Li, Hao Peng, Heng Ji.
MIT License
438 stars 36 forks source link

Potential performance issue: .fillna memory issue in pandas below 1.4.2 version #3

Closed TendouArisu closed 6 months ago

TendouArisu commented 6 months ago

Issue Description:

Hello. I have discovered a performance degradation in the .fillna function of pandas version 1.4.1 and below 1.4.2. And I notice the repository depends on pandas 1.4.1 in scripts/eval/science-world/requirements.txt. I am not sure whether this performance problem in pandas will affect this repository. I found some discussions on pandas GitHub related to this issue, including #46149 and #46204. I also found that scripts/data/interaction/collection/convert_outputs.ipynb and scripts/eval/mint-bench/convert_outputs.py used the influenced api. There may be more files using the influenced api and pandas version below 1.4.2.

Suggestion

I would recommend considering an upgrade to a different version of pandas >= 1.4.2 or exploring other solutions to optimize the performance. Any other workarounds or solutions would be greatly appreciated. Thank you!

xingyaoww commented 6 months ago

Hi @TendouArisu, thanks for the issue! I just removed the file scripts/eval/science-world/requirements.txt since we did not use it to setup environment (see here). The code should work fine if you use a newer version of pandas. Please let me know if you encountered more issues!