Data

Tools

Indexes

Applications

Experiments

Open Source Science

saev

Sparse autoencoders for vision

https://github.com/osu-nlp-group/saev

Science Score: 54.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Committers with academic emails
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (9.8%) to scientific vocabulary
Last synced: 7 months ago · JSON representation ·

Repository

Sparse autoencoders for vision

Basic Info
Statistics
  • Stars: 39
  • Watchers: 2
  • Forks: 6
  • Open Issues: 4
  • Releases: 0
Created over 1 year ago · Last pushed 7 months ago
Metadata Files
Readme Contributing License Citation Agents

README.md

saev - Sparse Auto-Encoders for Vision

Coverage

Sparse autoencoders (SAEs) for vision transformers (ViTs), implemented in PyTorch.

This is the codebase used for our preprint "Sparse Autoencoders for Scientifically Rigorous Interpretation of Vision Models"

About

saev is a package for training sparse autoencoders (SAEs) on vision transformers (ViTs) in PyTorch. It also includes an interactive webapp for looking through a trained SAE's features.

Originally forked from HugoFry who forked it from Joseph Bloom.

Read logbook.md for a detailed log of my thought process.

See related-work.md for a list of works training SAEs on vision models. Please open an issue or a PR if there is missing work.

Installation

Installation is supported with uv. saev will likely work with pure pip, conda, etc. but I will not formally support it.

Clone this repository, then from the root directory:

bash uv run python -m saev --help

This will create a virtual environment and display the CLI help.

Using saev

See the docs for an overview.

You can ask questions about this repo using the llms.txt file.

Example (macOS):

curl https://osu-nlp-group.github.io/saev/api/llms.txt | pbcopy, then paste into Claude or any LLM interface of your choice.

Owner

  • Name: OSU Natural Language Processing
  • Login: OSU-NLP-Group
  • Kind: organization
  • Location: United States of America

Citation (CITATION.cff) 

# This CITATION.cff file was generated with cffinit.
# Visit https://bit.ly/cffinit to generate yours today!

cff-version: 1.2.0
title: saev
message: >-
  If you use this software, please cite it using the
  metadata from this file.
type: software
authors:
  - given-names: Samuel
    family-names: Stevens
    email: samuel.robert.stevens@gmail.com
    orcid: 'https://orcid.org/0009-0000-9493-7766'
    affiliation: The Ohio State University
identifiers:
  - type: url
    value: 'https://arxiv.org/abs/2502.06755'
    description: Preprint
repository-code: 'https://github.com/OSU-NLP-Group/saev'
url: 'https://osu-nlp-group.github.io/saev/'
repository-artifact: 'https://pypi.org/project/saev/'
abstract: >-
  saev is a package for training sparse autoencoders (SAEs)
  on vision transformers (ViTs) in PyTorch. It also includes
  an interactive webapp for looking through a trained SAE's
  features.
keywords:
  - sparse autoencoders
  - interpretability
  - computer vision
license: CC-BY-4.0
commit: 5951e8497e16b06ebc5dbf38de0a87daf38331a5
date-released: '2025-04-28'

GitHub Events

Total
  • Issues event: 3
  • Watch event: 12
  • Issue comment event: 1
  • Push event: 45
  • Pull request event: 4
  • Gollum event: 3
  • Fork event: 3
  • Create event: 1
Last Year
  • Issues event: 3
  • Watch event: 12
  • Issue comment event: 1
  • Push event: 45
  • Pull request event: 4
  • Gollum event: 3
  • Fork event: 3
  • Create event: 1

Committers

Last synced: 9 months ago

All Time
  • Total Commits: 497
  • Total Committers: 6
  • Avg Commits per committer: 82.833
  • Development Distribution Score (DDS): 0.362
Past Year
  • Commits: 317
  • Committers: 1
  • Avg Commits per committer: 317.0
  • Development Distribution Score (DDS): 0.0
Top Committers
Name Email Commits
Samuel Stevens s****s@g****m 317
hugofry h****y@f****k 102
jbloom-md j****h@m****m 37
Joseph Bloom j****s@g****m 32
Lucy Farnik l****k@g****m 8
Slava Chalnev s****v@g****m 1
Committer Domains (Top 20 + Academic)
massdynamics.com: 1 fryfamily.co.uk: 1

Issues and Pull Requests

Last synced: 9 months ago

All Time
  • Total issues: 10
  • Total pull requests: 3
  • Average time to close issues: 2 days
  • Average time to close pull requests: about 2 hours
  • Total issue authors: 8
  • Total pull request authors: 3
  • Average comments per issue: 0.9
  • Average comments per pull request: 0.67
  • Merged pull requests: 1
  • Bot issues: 0
  • Bot pull requests: 0
Past Year
  • Issues: 10
  • Pull requests: 3
  • Average time to close issues: 2 days
  • Average time to close pull requests: about 2 hours
  • Issue authors: 8
  • Pull request authors: 3
  • Average comments per issue: 0.9
  • Average comments per pull request: 0.67
  • Merged pull requests: 1
  • Bot issues: 0
  • Bot pull requests: 0
View more stats
Top Authors
Issue Authors
  • samuelstevens (2)
  • isaac-w-dev (2)
  • johnbradley (1)
  • NielsRogge (1)
  • marcantoinewilly (1)
  • krishnakanthnakkav2 (1)
  • thompsonmj (1)
  • tambourine666 (1)
Pull Request Authors
  • samuelstevens (3)
  • isaac-w-dev (1)
  • marcantoinewilly (1)
  • jakesmbeattie (1)
Top Labels
Issue Labels
Pull Request Labels
codex (2)

Dependencies

pyproject.toml pypi
  • beartype >=0.19.0
  • datasets >=3.0.1
  • einops >=0.8.0
  • jaxtyping >=0.2.34
  • marimo >=0.9.10
  • pillow >=11.0.0
  • torch >=2.5.0
  • tqdm >=4.66.5
  • transformer-lens >=1.2.2
  • transformers >=4.45.2
  • tyro >=0.8.12
  • wandb >=0.18.5