salesforce/CodeRL

Question about the input of the critic model

Zyq-scut opened this issue · 1 comments

Hi, as you mention in the paper, the input of the critic model should include 'Ws' and 'D', where 'Ws' is the sampled program and 'D' is the problem description. However, I didn't find any operation that mixing 'Ws' to the input text in APPSBaseDataset or function generate_prompt. Is there anything I missed? If not, how would this difference affect the performance of the model?
Sincerely looking forward to your reply!

I got it now