ai-safety-foundation / sparse_autoencoder

Sparse Autoencoder for Mechanistic Interpretability
https://ai-safety-foundation.github.io/sparse_autoencoder/
MIT License
191 stars 39 forks source link

Fix incompatibility issue with PyTorch >= 2.2 #206

Closed HuFY-dev closed 8 months ago

HuFY-dev commented 8 months ago

Hard-coded params_t.