Language field type mismatch: str instead of list.
Hi, thanks much for your work in building this dataset.
I was trying to download a sample using:
pubmed_sample = load_dataset('HoangHa/pubmed25', split='train[:1%])
and got this error: "Field 'Language': Type mismatch. Target is <class 'str'>, Source is <class 'list'>. Skipping update."
I think it might be related to line 443 of pubmed25.py loading script:
"Language": datasets.Value("string")
I also noticed that it seems to download the entire dataset instead of a 1% sample.
rgds
stay
Ah yes, the current code downloads the entire dataset at once. You can modify it based on the files available here: https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/
TL;DR: the current code isn’t really UX-friendly—you’ll need to get your hands dirty at some point. If you want to investigate the subset of it, try streaming instead, check out: https://huggingface.co/docs/datasets/en/stream
By the way, the mismatch error is a known issue. It’s not critical, so I’ve left it as-is for now, but I do plan to fix it later. Thanks for the feedback!
k, will go study those. TQVM.