Spaces:

---

librarian-bots

/

tutorials

like 3

Running

Hey @davanstrien ! I've read your collection tutorial and have a few minor comments:

• Usage of notebook_login is (softly) deprecated in huggingface_hub as it's a simple alias for huggingface_hub.login(). So it's best to showcase login instead now. Exact same syntax and result but the advantage is that if someone copy-pastes it in a script, it still works :)
• instead of computing the 10% threshold and then cutting at 13 it's also possible to directly sort the list and then cut. Something like that should work:

datasets = sorted(datasets, key=lambda item: item.likes, reverse=1)[:math.ceil(len(datasets) * 0.10)]

The advantage is that you don't need the "get_threshold" method. The drawback is that datasets with an equal number of likes are separated in an arbitrary way.

• typo: The existed_ok parameter allows us to specify => the parameter is exists_ok (you have it right in the code snippet but typo in the text.
• also the last section shows the URL of the created collection but the link is broken (show https://huggingface.co/collections/librarian-bots/top-10-instruction-tuning-datasets-65117495134fd906b070c410 which gives error 400). Correct url seems to be https://huggingface.co/collections/librarian-bots/top-10-instruction-tuning-datasets-65117eeaca29f41ae7ae39fe.