Environments currently return role as part of content
Open
- Dominant language
- Python
- Stars
- 2k
- Forks
- 561
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 145
Description
https://github.com/NVIDIA-NeMo/RL/blob/0988a7dddb8c5896e28a40af1d36c63c869d2fa4/nemo_rl/environments/math_environment.py#L314-L317
This may have issues if someone uses the conversation format and apply_chat_template themselves since we may double apply the environment.
Contributor guide
Assessment
This issue has not been assessed yet.