/sparse_autoencoder

Sparse Autoencoder for Mechanistic Interpretability

Primary LanguagePythonMIT LicenseMIT

Sparse Autoencoder

PyPI PyPI - License Checks Release

A sparse autoencoder for mechanistic interpretability research.

pip install sparse_autoencoder

Demo

Check out the demo notebook for a guide to using this library.

Contributing

This project uses Poetry for dependency management, and PoeThePoet for scripts. After checking out the repo, we recommend setting poetry's config to create the .venv in the root directory (note this is a global setting) and then installing with the dev and demos dependencies.

poetry config virtualenvs.in-project true
poetry install --with dev,demos

Checks

For a full list of available commands (e.g. test or typecheck), run this in your terminal (assumes the venv is active already).

poe

Pre-Commit Hooks

You can install the pre-commit hooks, to automatically check for common issues before you commit, as follows:

poe precommit-setup