Add reproducibility model card to root

README.md

@@ -0,0 +1,45 @@
+---
+library_name: sae_lens
+tags:
+- sae
+- deepseek
+- qwen
+- interpretability
+---
+
+# Sparse Autoencoder (SAE) for deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
+
+This is a Sparse Autoencoder trained on layer 23 of `deepseek-ai/DeepSeek-R1-Distill-Qwen-7B`.
+
+## Reproducibility & Configurations
+- **Seed:** 42
+- **Architecture:** topk
+- **Expansion Factor:** 16
+- **Top-K:** 64
+- **Training Tokens:** 400000000
+- **Learning Rate:** 0.0003
+
+### Data
+- **Mixtures:** Smoltalk + OpenThoughts
+- **Total Prep Tokens:** 307322880
+- **Fraction Reasoning:** 0.2679 achieved vs 0.5 configured
+- **Fingerprint (SHA256):** `f474f0ff1226173b90aa88be3ae4dc8ea66a22f9e33b8e97709be70e959e50b8`
+
+## Evaluation Metrics (Held-out Split)
+- **L0:** 64.0
+- **FVU (Fraction of Variance Unexplained):** 1.150201410889183
+- **CE Loss (Clean):** 1.673828125
+- **CE Loss (Spliced SAE):** 10.943359375
+- **CE Loss (Zero Ablation):** 15.4091796875
+- **CE Recovered Fraction:** 0.32513330963384285
+
+## Validation / Usage
+
+```python
+from sae_lens import SAE
+from transformer_lens import HookedTransformer
+import torch
+
+model = HookedTransformer.from_pretrained_no_processing("deepseek-ai/DeepSeek-R1-Distill-Qwen-7B", dtype=torch.bfloat16)
+sae = SAE.from_pretrained("<YOUR_HF_USERNAME>/sae-r1-distill-qwen7b-l23", "qwen7b-distill-l23-topk64-x16-smoltalk_seed42/final_sae")
+```