Get Started with PyTorch

Simple examples to get started with using PyTorch, the deep learning framework.

PyTorch is an open source optimized tensor library and deep learning framework for Python. It is designed to be deeply integrated into Python.

PyTorch can be run inside the Urika-XCS image on a single compute node using Python multi-processing via the start_analytics command.

The PyTorch library consists of the following components:

Table 1. PyTorch Library Components
Component	Description
`torch`	Tensor library that provides both CPU and GPU support.
`torch.autograd`	Tape-based automatic differentiation library that supports all differentiable Tensor operations in Torch.
`torch.nn`	Neural networks library.
`torch.multiprocessing`	Used for Python multiprocessing.
`torch.utils`	DataLoader, Trainer and other utility functions.
`torch.legacy(.nn/.optim)`	Legacy code that has been ported over from Torch for backward compatibility reasons.
`torch.distributed`	Provides an MPI-like interface for exchanging tensor data across multiple nodes. It supports a few different backends and initialization methods.