---
license: apache-2.0
base_model: google/gemma-4-E2B
tags: [sparse-autoencoder, mechanistic-interpretability, sae-lens, gemma-4, batch-topk]
---

# gemma-4-e2b-scope-v1-L17-batchtopk-k64-seed17

BatchTopK SAE, L17, d_sae=24576, k=64, expansion 16x.
Training on FineWeb-Edu via Colab T4 bnb 4-bit inference.

**Progress:** 14,003,200 / 100,000,000 tokens (14.0 %). Steps: 13,675.

Live W&B: https://wandb.ai/caleb-deleeuw/gemma4-sae-scope/runs/gemma-4-e2b-scope-v1-L17-batchtopk-k64-seed17