Debug a simple PyTorch Lightning notebook

I created a PyTorch Lightning module that attempts to use transfer learning with a multi-headed version of EfficientNet. However, when calling the module's fit function, I receive a RuntimeError about a variable needed for gradient computation being modified by an inplace operation. I do not see where such an inplace operation occurs. I would like help identifying where this error occurs, so I can resolve it and ensure the model can train. I can provide more details and the notebook as needed.