---
language: en
library_name: mlsae
license: mit
tags:
- model_hub_mixin
- pytorch_model_hub_mixin
datasets:
- monology/pile-uncopyrighted
---

A Multi-Layer Sparse Autoencoder (MLSAE) trained on [EleutherAI/pythia-70m-deduped](https://huggingface.co/EleutherAI/pythia-70m-deduped) and [monology/pile-uncopyrighted](https://huggingface.co/datasets/monology/pile-uncopyrighted), with an expansion factor of 1 and $k = 16$.
For more details, see the [paper](https://arxiv.org/submit/5837813) and the [Weights & Biases project](https://wandb.ai/timlawson-/mlsae).

This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:

- Library: <https://github.com/tim-lawson/mlsae>
- Docs: <https://github.com/tim-lawson/mlsae/blob/main/mlsae/model/autoencoder.py>