TwT
TwT
Yes, I solved it. But it's been so many months that I've forgotten how I solved it. Can you try this version of trl to see if it can help...
Or refer to this stic/dpo_trainer.py for training. Sorry brother, I can't remember
Wow! Thanks a lot for your quick response!
I encountered a problem while performing inference following this document:https://github.com/princeton-nlp/SWE-bench/blob/main/swebench/inference/README.md When using the API model, the dataset filters out 226 instances for testing;  however, when using the LLaMA model,...
Has anyone solved this problem?