Why so many nans in the text-corpus?

#2
by djstrong - opened

Why so many nans in the text-corpus? Most of them are marked as relevant...

RAG Datasets org
edited Jun 16

Thanks!

I am getting the abstracts form pubmed via the get_abstract() in generate.py.
If I recall correctly some of the articles do not exist anymore which results in nans.

Would you have the capacity to check this and remove the nans form the text-corpus and the corresponding ids from the relevant_passage_ids?

We have fixed this by downloading missing abstracts: https://huggingface.co/datasets/enelpol/rag-mini-bioasq

thanks

Sign up or log in to comment