Mathux/TMR · Host models on Hugging Face

Very impressive demo! It seems the model is currently loaded from a Google Drive link. In general, we suggest to have the model weights on the Hub with a model card that documents usage. From the Space, you can use the huggingface_hub library to programmatically load model checkpoints from model repositories. This significantly helps with discoverability, documentation, and easy of usage.

Docs for uploading files manually: https://huggingface.co/docs/hub/models-uploading
Docs for programmatically downloading files https://huggingface.co/docs/huggingface_hub/guides/download