SDXL download too many files

[uzhao]

Simply download everything from stabilityai/stable-diffusion-xl-base-1.0 waste a lot of traffic.

For example, in unet folder there are
diffusion_flax_model.msgpack
10.3 GB
diffusion_pytorch_model.safetensors
10.3 GB
model.onnx_data
10.3 GB
openvino_model.bin
10.3 GB

If we can have a download function to prepare only necessary pretrain models in one step that would be very helpful.