Models that require CUDA don't work by default

[jeremyfowers]

Models that expect CUDA to be installed do not work by default on tkml.

Reproduction

turnkey onnxmodelzoo\toolchain\models\torch_hub\mealv1_resnest50.py

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Workaround

Installing torch+cuda with pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 solves the issue.

Possible solutions:

1. Add some logic to skip these models if CUDA is not installed, like if not torch.cuda.is_available(): return
2. See if the models can be fixed with something like model.device("cpu")
3. Remove models from the turnkey corpus that require CUDA

cc @danielholanda @ramkrishna2910 @andyluo7