GPU selection

[Siegurd01]

It would be great to add a feature in the ability to select a GPU on which the model will be deployed in systems with several GPUs. For example, I have a laptop where the AMD Radeon (4GB) video card is integrated, as well as the discrete Nvidia 3060 (6GB). Since the models are small, they can be deployed on the AMD video card, thus not interfering with the main video card which is used for other working purposes.
Thank you!

[Akuf]

I also have a multi GPU workstation and searching a way to load the Handy model on the GPU1 and not on the GPU0. Thanks in advance for any information.