How to load fine tuned model to huggingface?

[samarpan-rai]

I've fine tuned a huggingface model using flair for a classification task. I have it saved in a directory and and among many files, it has final-model.pt. I want to load this model to huggingface. How can I do this?

[stefan-it]

Hi @samarpan-rai ,

this is a very interesting question, because I'm now working on an integration of Hugging Face model hub with Flair.

For a sequence tagger model, an uploaded model (to HF model hub) can be used in Flair with this PR:

#1974

However, it is not yet supported for text classification models, but I could work on that next.

As Hugging Face is using git-lfs as storage, it will be awesome if you could upload your trained model under your desired namespace with a flair prefix, such as hf-user-name/flair-classification-dataset-name. That would be the repository name.

Your model should then be renamed to model.bin (because the .bin extension is already registered for lfs uploads).

Details about uploading a model with git-lfs are described here:

https://huggingface.co/transformers/model_sharing.html

After uploading the model, just ping me here on GitHub, then I'll do the Flair integration of the model loading 🤗

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.