Adding a dataset on hugging face?

[Sakshamrzt]

As stated here we might end up changing the original dataset to be compatible with the benchmark. In such case do we add the new dataset on huggingface under mteb or adding under our own user name is sufficient?

[KennethEnevoldsen]

You can supply your own user, in the future we might re-upload it under MTEB (as long as it has the right licenses) to ensure stability.

[Muennighoff]

Also, I'd say that everyone who contributes can also join the HF & GitHub orgs if they want to. I think on HF you can just send a request and then someone will add you as Contributor. Let me know if someone disagrees :)

[KennethEnevoldsen]

I think that is totally fine as well.

[Sakshamrzt]

Thanks for the quick answer!
However, the option to join the org on HF seems to be not activated. AFAIU the admin can enable this feature. @KennethEnevoldsen please enable if we are allowing contributors to join. Thanks!

[KennethEnevoldsen]

I have enabled the "requests to join".

[Sakshamrzt]

Thanks!

[rsgrewal-aws]

in addung your own data set -- the example provided does not work - it fails with pydantic error saying language en does not exist

[KennethEnevoldsen]

Hi @rsgrewal-aws, we recently changed to a new language standard to use the iso 639-3 language codes instead of the two letter codes (as these don't cover all languages). Additionally since languages can be in different scripts we also add a script code. Thus for english you would get "eng-Latn", English in Latin script. Hope it helps!