Hugging Face's logo

google
/
gemma-scope-2-270m-it

SAELens
Model card Files
xet
Community
CallumMcDougallGDM commited on
Commit
4490cee
·
verified ·
1 Parent(s): 7f32963

Add files using upload-large-folder tool

Browse files
Files changed (22) hide show
  1. resid_post_all/layer_10_width_262k_l0_small/config.json +3 -2
  2. resid_post_all/layer_11_width_262k_l0_big/config.json +3 -2
  3. resid_post_all/layer_12_width_262k_l0_small/config.json +3 -2
  4. resid_post_all/layer_13_width_262k_l0_big/config.json +3 -2
  5. resid_post_all/layer_13_width_262k_l0_small/config.json +3 -2
  6. resid_post_all/layer_14_width_16k_l0_small/config.json +3 -2
  7. resid_post_all/layer_14_width_262k_l0_big/config.json +3 -2
  8. resid_post_all/layer_16_width_16k_l0_big/config.json +3 -2
  9. resid_post_all/layer_16_width_262k_l0_big/config.json +3 -2
  10. resid_post_all/layer_1_width_16k_l0_big/config.json +3 -2
  11. resid_post_all/layer_1_width_262k_l0_big/config.json +3 -2
  12. resid_post_all/layer_6_width_16k_l0_big/config.json +3 -2
  13. resid_post_all/layer_6_width_262k_l0_big/config.json +3 -2
  14. resid_post_all/layer_7_width_16k_l0_small/config.json +3 -2
  15. resid_post_all/layer_7_width_262k_l0_big/config.json +3 -2
  16. resid_post_all/layer_8_width_16k_l0_small/config.json +3 -2
  17. resid_post_all/layer_8_width_262k_l0_small/config.json +3 -2
  18. resid_post_all/layer_9_width_16k_l0_big/config.json +3 -2
  19. resid_post_all/layer_9_width_262k_l0_small/config.json +3 -2
  20. transcoder/layer_9_width_16k_l0_medium/config.json +3 -2
  21. transcoder/layer_9_width_16k_l0_small/config.json +3 -2
  22. transcoder/layer_9_width_16k_l0_small_affine/config.json +3 -2
resid_post_all/layer_10_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.output",
3
  "hf_hook_point_out": "model.layers.10.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.output",
3
  "hf_hook_point_out": "model.layers.10.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_11_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.output",
3
  "hf_hook_point_out": "model.layers.11.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.11.output",
3
  "hf_hook_point_out": "model.layers.11.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_12_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.12.output",
3
  "hf_hook_point_out": "model.layers.12.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.12.output",
3
  "hf_hook_point_out": "model.layers.12.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_13_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.13.output",
3
  "hf_hook_point_out": "model.layers.13.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.13.output",
3
  "hf_hook_point_out": "model.layers.13.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_13_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.13.output",
3
  "hf_hook_point_out": "model.layers.13.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.13.output",
3
  "hf_hook_point_out": "model.layers.13.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_14_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.14.output",
3
  "hf_hook_point_out": "model.layers.14.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.14.output",
3
  "hf_hook_point_out": "model.layers.14.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_14_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.14.output",
3
  "hf_hook_point_out": "model.layers.14.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.14.output",
3
  "hf_hook_point_out": "model.layers.14.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_16_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.16.output",
3
  "hf_hook_point_out": "model.layers.16.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.16.output",
3
  "hf_hook_point_out": "model.layers.16.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_16_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.16.output",
3
  "hf_hook_point_out": "model.layers.16.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.16.output",
3
  "hf_hook_point_out": "model.layers.16.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_1_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.1.output",
3
  "hf_hook_point_out": "model.layers.1.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.1.output",
3
  "hf_hook_point_out": "model.layers.1.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_1_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.1.output",
3
  "hf_hook_point_out": "model.layers.1.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.1.output",
3
  "hf_hook_point_out": "model.layers.1.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_6_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.6.output",
3
  "hf_hook_point_out": "model.layers.6.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.6.output",
3
  "hf_hook_point_out": "model.layers.6.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_6_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.6.output",
3
  "hf_hook_point_out": "model.layers.6.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.6.output",
3
  "hf_hook_point_out": "model.layers.6.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_7_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.7.output",
3
  "hf_hook_point_out": "model.layers.7.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.7.output",
3
  "hf_hook_point_out": "model.layers.7.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_7_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.7.output",
3
  "hf_hook_point_out": "model.layers.7.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.7.output",
3
  "hf_hook_point_out": "model.layers.7.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_8_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.8.output",
3
  "hf_hook_point_out": "model.layers.8.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.8.output",
3
  "hf_hook_point_out": "model.layers.8.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_8_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.8.output",
3
  "hf_hook_point_out": "model.layers.8.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.8.output",
3
  "hf_hook_point_out": "model.layers.8.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_9_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.output",
3
  "hf_hook_point_out": "model.layers.9.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.output",
3
  "hf_hook_point_out": "model.layers.9.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
resid_post_all/layer_9_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.output",
3
  "hf_hook_point_out": "model.layers.9.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.output",
3
  "hf_hook_point_out": "model.layers.9.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
transcoder/layer_9_width_16k_l0_medium/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "transcoder"
10
  }
transcoder/layer_9_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "transcoder"
10
  }
transcoder/layer_9_width_16k_l0_small_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-it",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }