Reverse Dependencies of transformer-lens

The following projects have a declared dependency on transformer-lens:

• auto-circuit — no summary
• Auto-HookPoint — Make any model compatible with transformer_lens
• belief-state-superposition — Investigating belief state representations of transformers trained on Hidden Markov Model emissions
• codebook-features — Sparse and discrete interpretability tool for neural networks
• devinterp — A library for doing research on developmental interpretability
• e2e-sae — Repo for training sparse autoencoders end-to-end
• feature-lens — In-depth visualizations for SAE features
• graphpatch — graphpatch is a library for activation patching on PyTorch neural network models.
• hugginglens — no summary
• mamba-lens — TransformerLens port for Mamba
• mistral-sae — Sparse AutoEncoder to decode Mistral LLM
• pattern-lens — no summary
• reverseAbliterator — A package for reverse ablitation of language models
• sae-dashboard — Open-source SAE visualizer, based on Anthropic's published visualizer. Forked / Detached from sae_vis.
• sae-lens — Training and Analyzing Sparse Autoencoders (SAEs)
• sae-vis — Open-source SAE visualizer, based on Anthropic's published visualizer.
• smol-sae — Minimal implementation of SAEs
• token-trace — Transformer token flow visualizer
• transcoders-slim — A template for python projects in PDM
• transpector — Visually inspect, analyse and debug transformer models. Aimed at reducing cycle times for interpretability research and lowering the barrier to entry.
• tuned-lens — Tools for understanding how transformer predictions are built layer-by-layer