[Feature]: Allow to load models locally from a file path

[vrdn-23]

What feature would you like to request?

Hi,
Apologies if this can be already done, but I'm new to using fastembed and I was wondering if it would be possible to load the model files for a particular embedding model through a local user-defined directory instead of having to go though the huggingface download pipeline each time. Providing a local path directly currently throws an error since it does not appear in the list of supported models.

Is there any additional information you would like to provide?

A useful scenario for this is where we are pulling the model weights from elsewhere and running this in a Kubernetes pod. Since pods are ephemeral by nature, having them re-pull the model weights each time if they're not found in the cache seems tedious and unnecessary. I'd be happy to help with a PR towards this if required.

Apologies again if this is already possible and I somehow missed this! Thanks again for the awesome project!

[vrdn-23]

I can see that a duplicate issue #321 has already been raised for this, but I would like to keep this open because the new one seems inactive and without maintainer response.

@joein is this something the dev team would like some help with? I would be happy to volunteer my time and effort

[osehmathias]

Agree. This would be amazing.