Fastest way to upload custom TensorFlow model/weights?

Hi,

Most of the tutorials I have seen focus on loading existing large models. I have a custom model + saved weights. How can I upload this to hugging face hub?

I would like to firstly load the model in a notebook. Ultimately I want to use this model in a gradio demonstration.

Thanks!

I’m assuming you’re talking about TF models in the transformers library.

You can use the push_to_hub method or the PushToHubCallback callback to share a model in the Hub. There is a guide for it on Share a model that might be interesting to you. For custom models Sharing custom models might be of interest too.

1 Like

cc @Rocketknight1 and @merve if they want to add anything else.