Private Huggingface models on Model Garden

[d1-fast]

Is it possible to deploy a custom HF model on model garden. I see there is an option to "Deploy from Hugging Face" however I get an error when trying to provide an access token my private HF model. I found this thread (https://discuss.huggingface.co/t/vertex-ai-the-token-is-not-valid-or-not-have-permission/91101/7) which is similar but it seems to be trying to deploy LLaMA. I am interested in deploying a model that I've trained and uploaded to HF. Is this possible with model garden?

[McMaco]

Hello d1-fast,

Welcome to Google Cloud Community!

You’re right! It is possible to deploy a custom model from Hugging Face but before deploying we must create a token to access private models on Hugging Face Hub.

I hope the above information is helpful.

[d1-fast]

I've created a token and tried the most broad permissions and still get the permissions issue. In fact, the UI requests a HF token, so you can't make progress on the console without providing one.

[McMaco]

You may use this documentation for further information in using hugging face model for deployment.

I hope it helps.

[si_gcp]

In the documentation you've provided, I don't see a way to deploy a private model to HF. The docs only give you a guide for public, open-source models.