env issue

[jiaweixu98]

Thank you for open-sourcing everything. I tried to run a quick test on the inference, but it failed. It looks like the environment dependencies are tricky. Some additional work may be needed to tweak the code to make it easier to reproduce.

Thanks,
Jiawei

[jiaweixu98]

I have tried multiple times. The MHQA agent works as expected, but the code agent (7B-RL model) does not, likely due to some minor configuration issues. I would greatly appreciate it if the authors could verify the setup and ensure it works by following the instructions in the README.

[jiaweixu98]

Finally, it works! However, I still think it would be much easier for people to replicate if we better organize the environment dependencies and instructions.