Implementing and training a model in PyTorch
PyTorch is a Python library for Torch, a ML package for Lua. The main features of PyTorch include graphics processing unit- (GPU-) accelerated matrix calculation and automatic differentiation for building and training neural networks. Creating the computation graph dynamically as the code gets executed, PyTorch is gaining popularity for its flexibility and ease of use, as well as its efficiency in model training.
Built on top of PyTorch, PyTorch Lightning (PL) provides another layer of abstraction, hiding many boilerplate codes. The new framework pays more attention to researchers by decoupling research-related components of PyTorch from the engineering-related components. PL codes are typically more scalable and easier to read than PyTorch codes. Even though the code snippets in this book put more emphasis on PL, PyTorch and PL share a lot of functionalities, so most components are interchangeable. If you are willing to dig into the...
