chapter 1 - introducing deep learning and the PyTorch library

• torch.nn contains the common neural network layers and components
  • connected layers, convolutional layers, activation functions, and loss functions
• we need to convert each sample from our data into something PyTorch can actually handle: tensors
  • The bridge between our custom data and a standardized tensor is the Dataset class in torch.utils.data
  • we'll go over this in chap 4
  • data storage is often slow, so we want to parallelize data loading
    • in order to assemble the data into batches (tensors that encompass several samples)
    • the DataLoader class handles this automatically
    • cf chap 7
• at each step of the training loop, we evaluate our model on the samples we got from the data loader
  • we then compare the outputs of our model to the desired output using a criterion or loss function
    • provided in torch.nn
  • to push the model to resemble the target, we use an optimizer
    • note: what does "autograd" mean? auto gradient?
• PyTorch provides a way to compile models ahead of time through TorchScript
  • serializes a model into a set of instructions that can be invoked independently of Python, from C++ programs or mobile devices
  • a model can be exported as torchscript or a standard format called ONNX
  • for deployment
  • cf chap 15