In this recipe, we used a Docker container from NVIDIA to bypass the many steps it requires to install NVIDIA GPU on a local computer. We used VS Code to connect to the running Docker container and we tested it to make sure the container was capable of using the GPUs. We then developed our code.
First, as always, we imported our libraries. Then we declared our variables. The first variable is the location of the training data, the split amount, the number of epochs, and the steps run. We then made a function that prints the results on screen so that we could see whether our model was improving with changes to the hyperparameters. We then imported the images from our training folder. After that, we set up our neural network. Next, we imported the ResNet 50 model. We set the model's requires_grad parameters to false so that our code would not affect the already existing model. We are using a sequential linear neural network using ReLU for our activation function with...