LFM2.5-230M SAEs (layers 0–13)

Sparse autoencoders for LiquidAI/LFM2.5-230M, one SAE per residual-stream layer.

Layout

l0/
  sae.pt
  norm.pt
  umap.json
l1/
  ...
l13/
  sae.pt
  norm.pt
  umap.json

Specs

Field Value
Base model LiquidAI/LFM2.5-230M
Layers 0–13 (14 SAEs)
Files per layer sae.pt (~256 MiB), norm.pt, umap.json
Total package size ~3.6 GiB

Files

  • sae.pt — SAE checkpoint for that layer
  • norm.pt — activation normalization stats
  • umap.json — feature UMAP coordinates for visualization

License

  • This SAE release: Apache License 2.0
  • Base model: LFM Open License v1.0 (Liquid AI). Using these SAEs with LFM2.5-230M remains subject to that license.

Acknowledgements

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Aquinlabs/lfm2.5-230m-sae

Finetuned
(28)
this model