docs: update datasets tutorial

[SauravMaheshkar]

This PR aims to update and refactor Tutorial 1 "Structure your Input"

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.81%. Comparing base (cbd5495) to head (f932e71).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1569   +/-   ##
=======================================
  Coverage   81.81%   81.81%           
=======================================
  Files         144      144           
  Lines        6120     6120           
=======================================
  Hits         5007     5007           
  Misses       1113     1113           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MalteEbner]

Looks good, I like it!

What I am most unsure about: Did you try actually feeding a hf_dataloader into the Lightly SSL training scripts? They might be incompatible, as the Lightly SSL training scripts expect a (batch, targets, filenames) dataloader output, which is not provided by huggingface.

[SauravMaheshkar]

g scripts? They might be incompatible, as the Lightly SSL training scripts expect a (batch, targets, filenames) dataloader output, which is not provided by huggingface.

Yes, I've tried it out and it does work with Lightly Training Scripts. The Colab Notebook released with the recent vector index blogpost uses the food101 dataset from the Huggingface for training.

[MalteEbner]

Looks good from my side :)

[guarin]

Thanks for adding all this helpful information! I left a bunch of comments, mostly regarding structure and formatting.

[guarin]

Looks great! Two small things left and then we can merge :)

[guarin]

LGTM! Thanks a lot!