Why does training require Tensorflow?

[VisionaryMind]

I have PyTorch 1.11.0 with CUDA 10.2 installed, but when attempting to run training, the process dies due to a lack of a Tensorflow dependency. Where should the configuration be changed to ensure PyTorch is used. Perhaps this code does not work with PyTorch 1.11?

[VisionaryMind]

For anyone that encounters this problem, you can resolve by downgrading to these versions in the dependency stack:

scipy==1.2.0
torch==1.9.0+cu111
torchvision==0.10.0+cu111
tensorflow==1.15.0