search

RelativelyFine / Latent-Feature-Interpretation-SAE

Executing Brain Surgery on Neural Networks.
MIT License
0 stars 0 forks source link

Implement a Toy Model #7

Open RelativelyFine opened 1 week ago

RelativelyFine commented 1 week ago

Recreate a the 1 layer transformer in this paper: https://transformer-circuits.pub/2023/monosemantic-features/index.html

This library may help: https://github.com/shehper/sparse-dictionary-learning

RelativelyFine commented 3 days ago

Teams: @Brauch25 - @dkrayacich @DhruvPopli - @santosrojella

use vscode liveshare to collaborate set up meeting times in the discord channels