PubMedQA Finetuning

Open mahmouds-cerebras opened this issue 2 years ago • 1 comments

I'm trying to replicate the fine-tuning on PubMedQA and I'm getting different results compared to the published Model and the numbers reported in the paper. I fine-tuned the pertained model directly on pqal data which I believe might be the reason, so did you do multi-phase fine-tuning for PubMedQA on pqa-A, pqa-U, pqa-L? could you please provide the hyperparams of each phase? Thanks.

May 29 '23 17:05 mahmouds-cerebras

@mahmouds-cerebras Can you please share your code on how you are fine-tuning the model?

Jun 04 '23 21:06 upadhayay