It appears that the get_custom_reward_fn function in main_ppo.py is redundant as there is a similar function defined in trainer/ppo/reward.py. Since both functions serve the same purpose, the one in main_ppo.py should be considered for removal to avoid duplication and maintain cleaner code.
https://github.com/volcengine/verl/blob/54b2677f72f6720f80480d72af7f47eee74b4dc2/verl/trainer/main_ppo.py#L27-L59
and
https://github.com/volcengine/verl/blob/54b2677f72f6720f80480d72af7f47eee74b4dc2/verl/trainer/ppo/reward.py#L25-L57
Proposed Action:

• Review the implementation of get_custom_reward_fn in both files.

• If they are indeed identical or serve the same functionality, remove the one in main_ppo.py.

Thank you!