Inferno Documentation

Inferno is an extensible library for simulating spiking neural networks. Built on top of PyTorch, it is designed with machine learning practitioners in mind. This project is still an early release and features may be subject to change.

Installation

At this time, installation is only supported via pip. Future support for installation with conda is planned.

With PyTorch (CPU Only)

pip install inferno-ai[torch]

Installation with the “torch” flag will install PyTorch 2.4.1 and the corresponding version of TorchVision.

Without PyTorch

pip install inferno-ai

Note: Inferno still requires PyTorch and a version of it must be installed. PyTorch will need to be installed separately if CUDA support is required. See the instructions for installing PyTorch for details.

Getting Started

For an example of walking through building, training, and evaluating a spiking neural network, see the quickstart example. A comprehensive documentation of the classes and functions provided by Inferno are available here.