Giters

CarperAI / trlx

A repo for distributed training of language models with Reinforcement Learning via Human Feedback (RLHF)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Implements beam search of AutoModelForCausalLMWithILQLHeads

zwhe99 opened this issue · comments

commented

🚀 The feature, motivation, and pitch

I notice that AutoModelForCausalLMWithILQLHeads has its own generate function, which only supports sampling. Would it be possible to use beam search?

Alternatives

No response

Additional context

No response

commented

We definitely agree this would be a useful addition, however there are other features we are prioritising right now. It would be a nice pr if you feel like addressing this yourself!