fix: align readme with current mteb

[Samoed]

Checklist

• Run tests locally to make sure nothing is broken using make test.
• Run the formatter to format the code using make lint.

The current README is a bit misleading and causing issues for newcomers. See #1490 and #1491 for reference.

[KennethEnevoldsen]

Should we just recommend this one always?

and maybe rephrase to:

Suggested change

	model = mteb.get_model(model_name)
	model = mteb.get_model(model_name) # if the model is not implemented in MTEB it will be eq. to SentenceTransformer(model_name)

[Samoed]

Yes, but I'm not sure removing it is the best approach. I can make changes to whatever you think is better

[KennethEnevoldsen]

Why does it need to inherit from wrapper? I would instead inherit from the Encoder protocol

[Samoed]

MTEB class will wrap models that do not inherit from the wrapper class.

mteb/mteb/evaluation/MTEB.py

Lines 366 to 367 in 3ff38ec

	if not isinstance(model, Wrapper):
	model = SentenceTransformerWrapper(model)

[KennethEnevoldsen]

Hmm, shouldn't it just wrap SentenceTransformers?

[Samoed]

It can be like this. You changed it in mieb

mteb/mteb/evaluation/MTEB.py

Lines 356 to 357 in fab0b82

	if isinstance(model, (SentenceTransformer, CrossEncoder)):
	model = SentenceTransformerWrapper(model)

I can change there too

[KennethEnevoldsen]

Hmm then it will also be merged into v2.0.0 in which case we should probably just update the readme there (I plan to merge it during December)

[Samoed]

So, for now I think this PR can be merged, and I will update readme in the 2.0 branch