[docs] create a diagram image that explains the hf:// paths

[severo]

On the same type of https://huggingface.co/docs/hub/gguf

I also did something similar for the converted Parquet files: https://www.linkedin.com/feed/update/urn:li:activity:7092975830020075520

Figma here: https://www.figma.com/file/Njhkk9wVE5E9tYlXrpSjo5/ParquetURL?type=design&node-id=0%3A1&mode=design&t=m82D8BqyKh2jhtak-1

[severo]

cc @julien-c @lhoestq (https://huggingface.slack.com/archives/C04BP5S7858/p1710583335439119)

[julien-c]

the parquet one looks nice!

[Wauplin]

Nice idea, yes!

[julien-c]

maybe re-do the parquet one but for DuckDB hf:// paths and display it inside the DuckDB related docs?

[severo]

wdyt?

https://www.figma.com/design/D0S8By7C8Use8ZDdVZZj8P/hf%3A%2F%2F-URLs

[julien-c]

maybe do a branch example with a simpler branch name like @main?

[julien-c]

and i think i would remove spaces and models to keep things focused on datasets

[julien-c]

and mention that refs/convert/parquet is auto-conversion

[julien-c]

cc @cfahlgren1 might be interesting to you!

[cfahlgren1]

This is great, maybe for the parquet branch, a mention of partial conversions up to 5gb might be helpful.

[cfahlgren1]

Not sure if it would clutter the graphic, but an auth example is good to know for users querying their private datasets.

CREATE SECRET hf_token (
   TYPE HUGGINGFACE,
   TOKEN 'your_hf_token'
);

[severo]

OK. Version A, for huggingface_hub hf:// docs:

[severo]

Version B, for duckdb

[julien-c]

neat!

[severo]

huggingface/hub-docs#1319 and #2358

[severo]

https://huggingface.co/docs/hub/datasets-duckdb
https://huggingface.co/docs/huggingface_hub/main/en/guides/hf_file_system#integrations