Update README.md
Browse files
README.md
CHANGED
@@ -1,5 +1,31 @@
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
|
|
3 |
---
|
4 |
|
5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
3 |
+
library_name: saelens
|
4 |
---
|
5 |
|
6 |
+
⚠️ WARNING: we are still uploading these SAEs
|
7 |
+
|
8 |
+
# 1. Gemma Scope
|
9 |
+
|
10 |
+
Gemma Scope is a comprehensive, open suite of sparse autoencoders for Gemma 2 9B and 2B. Sparse Autoencoders are a "microscope" of sorts that can help us break down a model’s internal activations into the underlying concepts, just as biologists use microscopes to study the individual cells of plants and animals.
|
11 |
+
|
12 |
+
See our [landing page](https://huggingface.co/google/gemma-scope) for details on the whole suite. This is a specific set of SAEs:
|
13 |
+
|
14 |
+
# 2. What Is `gemma-scope-2b-pt-transcoders`?
|
15 |
+
|
16 |
+
- `gemma-scope-`: See 1.
|
17 |
+
- `2b-pt-`: These SAEs were trained on Gemma v2 2B base model.
|
18 |
+
- `transcoders`: These SAEs are transcoders: they were trained to reconstruct the output of MLP sublayers from the input to the MLP sublayers: see https://arxiv.org/abs/2406.11944
|
19 |
+
|
20 |
+
# 3. Point of Contact
|
21 |
+
|
22 |
+
Point of contact: Arthur Conmy
|
23 |
+
|
24 |
+
Contact by email:
|
25 |
+
|
26 |
+
```python
|
27 |
+
''.join(list('moc.elgoog@ymnoc')[::-1])
|
28 |
+
```
|
29 |
+
|
30 |
+
HuggingFace account:
|
31 |
+
https://huggingface.co/ArthurConmyGDM
|