Managed to pass all of our tests locally on my Mac :).

Some functions work with MPS support, whereas others don't. For the ones that don't, I elected to just pass to the calculations to the CPU.

1. The model Faster_RCNN fails because of non-implemented pytorch function, as well as the MobileSAMv2 model (aten::upsample_bicubic2d.out)
2. SAM_Automatic_Mask_Generation fails because of floating point transformations. I think this is just caused by numpy defaulting to 64-bit precision, because SAM prediction works when we're just working with one mask at a time.

But what works is the YOLOv8 bounding box prediction, which is a pretty big step in our model.

Also, updating license to include mobileSAMv2 and the Cellpose dataset.

Partially resolves #5