How can we use the sam-vit-huge in the production?

[moneyhotspring]

Question

The size of ONNX files for sam-vit-huge is around 600MB. If I am using the implementation mentioned in the documentation, it downloads these files first before performing the image segmentation. Is there a better way to avoid downloading these files and reduce the time it takes? Additionally, the model is taking too much time to generate embeddings when using sam-vit-huge or sam-vit-large.