[inference] Necessary breaking change: nest task-specific route inside of model route

[julien-c]

The alternative is just to break this feature...

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.00%. Comparing base (37ab5ba) to head (5fdf61e).
Report is 17 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #3044       +/-   ##
===========================================
+ Coverage   43.75%   81.00%   +37.24%     
===========================================
  Files         124      124               
  Lines       12261    12430      +169     
===========================================
+ Hits         5365    10069     +4704     
+ Misses       6896     2361     -4535     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[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.

[hanouticelina]

Just to make sure I understand from the context of this internal thread: the goal here is to point feature-extraction and sentence-similarity tasks to the same model route so that the same TEI deployment can answer both kinds of calls, right? (instead of having two deployments of the exact same model under its own /pipeline).
The PR looks good but let's wait for the update server-side (iiuc the PR won't work without switching to the TEI container).

[julien-c]

the goal here is to point feature-extraction and sentence-similarity tasks to the same model route so that the same TEI deployment can answer both kinds of calls

not necessarily TEI, but yeah, the same container would be able to answer both calls/tasks, unlike now where we need two (redundant) running containers (unless we'd make complex URL/route rewriting)

Can you confirm this @oOraph?

[hanouticelina]

I added unit tests and tested the PR 👍 all good, let's merge this (the inference tests are passing).

from huggingface_hub import InferenceClient


client = InferenceClient(provider="hf-inference")

similarities = client.sentence_similarity(
    model="intfloat/e5-base-v2",
    sentence="Hello, how are you?",
    other_sentences=["Hello, how are you?", "I am fine, thank you!"],
)

print(similarities)