DavidUdell / sparse_circuit_discovery

Circuit discovery in GPT-2 small, using sparse autoencoding
MIT License
6 stars 1 forks source link

Feature/web #13

Closed DavidUdell closed 9 months ago

DavidUdell commented 9 months ago

feature_web now yields causal graphs for sparse autoencoder features, both on RASP toy models from tracr and on full-scale HF transformers.