Input/completion in reward training vs. RL training

[menandro]

Why is it in reward training, the input and completion is appended as: user_input + " " + completion (reward.py line 254) where as in RL training, the equivalent task_response is: input + "\n" + completion (trainer.py line 680)?

[diegofiori]

Hi @menandro, thank you for reporting the mismatch! I think it's just a typo that we need to fix as soon as possible. Would you mind opening a PR to fix this problem? I think it makes much more sense to use the formula we use in reward.py.