detect_from_video.py error when running on MesoNet Model

[Kavilan Nair]

There is a bug in the codebase that prevents a user from running inference using the MesoNet model.

 python detect_from_video.py -i deepfake_videos -mi faceforensics++_models_subset/face_detection/Meso/Meso4_deepfake.pkl -mt meso -o meso_deepfakes/ --cuda

Provides the following error:

 File "detect_from_video.py", line 289, in <module>
    test_full_image_network(**vars(args))
  File "detect_from_video.py", line 156, in test_full_image_network
    model = model.cuda()
AttributeError: 'collections.OrderedDict' object has no attribute 'cuda'
  0%|                                                        | 0/10 [00:03<?, ?it/s]