Skip to content

About

Abliterator (llm behavior/concept removal) with baukit, not transformerlens

Resources

Stars

11 stars

Watchers

1 watching

Forks

Latest commit

 

History

23 Commits

Folders and files

Repository files navigation

wassname's Abliterator (with baukit not transformerlens)

What is this? It's a concept removal method, here I show how to use baukit, which is a nice library which uses pytorch hooks to modify layers at runtime.

setup

uv sync

then open nbs/02b_abliterate_lr.ipynb in Jupyter Notebook.

About

Abliterator (llm behavior/concept removal) with baukit, not transformerlens

Resources

Stars

11 stars

Watchers

1 watching

Forks

Releases

Packages

Used by

Contributors

Languages