Multi-Layer Sparse Autoencoders
Collection
Single SAEs trained on the residual stream activation vectors from every transformer layer simultaneously: https://arxiv.org/abs/2409.04185
•
30 items
•
Updated
mlsae-pythia-70m-deduped-x1-k32
A Multi-Layer Sparse Autoencoder (MLSAE) trained on the residual stream activation vectors from every layer of EleutherAI/pythia-70m-deduped with an expansion factor of 1 and k = 32, over 1 billion tokens from monology/pile-uncopyrighted.
For more details, see:
- Paper: https://arxiv.org/abs/2409.04185
- GitHub repository: https://github.com/tim-lawson/mlsae
- Weights & Biases project: https://wandb.ai/timlawson-/mlsae
- Downloads last month
- 8