search

facebookresearch / DPR

Dense Passage Retriever - is a set of tools and models for open domain Q&A task.
Other
1.71k stars 300 forks source link

question: reader loss #244

Open PaulLerner opened 1 year ago

PaulLerner commented 1 year ago

Hi,

I have trouble understanding this line in compute_loss in reader.py: https://github.com/facebookresearch/DPR/blob/a31212dc0a54dfa85d8bfa01e1669f149ac832b7/dpr/models/reader.py#L109

This keeps the maximum loss over all M passages, why? Why not summing or averaging?

Best regards,