Fix URL in get_safetensors_metadata docstring

[Wauplin]

Tweak safetensors parser: fetch only 25kb by default (instead of 100kb currently). Still valid for most use cases (82% in top 1000 models) and save some bandwidth. Most probably won't be noticed. (see #1855 (comment)) => EDIT: not changed in the end.

Fix typo in docs: wrong url in get_safetensors_metadata docstring.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (2bdab6a) 82.42% compared to head (d76e293) 82.48%.

❗ Current head d76e293 differs from pull request most recent head 9ec3206. Consider uploading reports for the commit 9ec3206 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1951      +/-   ##
==========================================
+ Coverage   82.42%   82.48%   +0.06%     
==========================================
  Files          66       66              
  Lines        8154     8155       +1     
==========================================
+ Hits         6721     6727       +6     
+ Misses       1433     1428       -5     

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

[LysandreJik]

Did we encounter a problem with 100kb? I think it's reasonable enough already so wondering if we need the change

[Wauplin]

Did we encounter a problem with 100kb? I think it's reasonable enough already so wondering if we need the change

No problems reported no.
It's just that I thought saving 4x bandwidth in 82% percent of cases while adding an extra HEAD call in 15% of cases seemed reasonable. But fine with keeping 100kb if you feel like it

[julien-c]

i would also tend to not change anything, personally..

[Wauplin]

Ok, reverted in 456555a.

PR becomes "fix typo in docs" :)