Placing model on GPU

[enzokro]

Hello! Thank you for the incredible models. At a first glance, it seems like the Predictors are always on the CPU by default, and I can't find an easy way in the docling* libraries to place these models on an accelerator (i.e. MPS or Cuda).

What would be the team's suggested way of using GPUs with the models? Thanks in advance for the help!