Introduction to functional PyTorch

How to use write Jax-style PyTorch models

PyTorch has recently integrated the torch.func module into its main codebase in the 2.0 release. This module, previously known as functorch, enables the development of purely functional neural network models in PyTorch with a straightforward API. This package is PyTorch’s response to the growing popularity of Jax, a Python framework for general differentiable programming built using a functional programming paradigm from the ground up.

In …

api codebase development functional-programming introduction jax machine learning network neural network neural networks package pytorch release