dag-saebench-layer12-aegyekpc

DAG Model for saebench SAE

This repository contains a trained Directed Acyclic Graph (DAG) model for measuring effective L0 of a Sparse Autoencoder.

Model Info

• SAE Type: saebench
• SAE Release: canrager/saebench_gemma-2-2b_width-2pow14_date-0107
• SAE ID: gemma-2-2b_jump_relu_width-2pow14_date-0107/resid_post_layer_12/trainer_1
• d_sae: 16384
• Tokens Used: 10,000,000
• Effective L0: 22
• Actual L0: 39.9
• Compression Ratio: 1.81x

Files

• final_model.safetensors: Trained DAG model (Lambda matrix, b_penalty, feature_order)
• results.json: Training metadata and metrics
• training_curves.png: Loss curves and training progress visualization

Usage

Use with the Probabilistic SAE Streamlit dashboard:

1. Check "Load pre-trained DAG from HF"
2. DAG model HF repo: TheodoreEhrenborg/dag-saebench-layer12-aegyekpc
3. DAG model subfolder: (leave empty)

The dashboard will automatically load the matching SAE and enable clustering.

Training Details

Trained using effective_l0_vanilla.py with:

• Epochs: 1
• Learning rate: 0.0005
• Batch size: 6400

For more details, see results.json.