Skip to main content

Install

Write a model

This is a two-layer spiking network: 700 input features → 512 hidden LIF neurons → 20 output LIF neurons. It matches the architecture used in the keyword-spotting tutorial.

Compile

model.thx is a sealed artifact containing the architecture, weights, and resolved runtime constants (leak factor alpha, CRC32 integrity hash). It is self-contained — no Python environment needed to run it.

Run

The transcript prints to stdout. It is structured, machine-readable, and snapshot-tested — the format is a contract.

Next steps

First model

Build, train, and run a model end to end with real data.

Keyword spotting tutorial

Full training run on the Spiking Heidelberg Digits dataset.