Hugging Face's logo

google
/
gemma-scope-2-12b-pt

SAELens
Model card Files
xet
Community
CallumMcDougallGDM commited on
Commit
8d55106
·
verified ·
1 Parent(s): c1d2705

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. attn_out_all/layer_0_width_262k_l0_small/config.json +3 -2
  2. attn_out_all/layer_10_width_16k_l0_small/config.json +3 -2
  3. attn_out_all/layer_10_width_262k_l0_small/config.json +3 -2
  4. attn_out_all/layer_12_width_262k_l0_small/config.json +3 -2
  5. attn_out_all/layer_13_width_16k_l0_small/config.json +3 -2
  6. attn_out_all/layer_14_width_16k_l0_big/config.json +3 -2
  7. attn_out_all/layer_15_width_16k_l0_small/config.json +3 -2
  8. attn_out_all/layer_16_width_16k_l0_small/config.json +3 -2
  9. attn_out_all/layer_16_width_262k_l0_big/config.json +3 -2
  10. attn_out_all/layer_17_width_16k_l0_big/config.json +3 -2
  11. attn_out_all/layer_19_width_262k_l0_big/config.json +3 -2
  12. attn_out_all/layer_1_width_262k_l0_big/config.json +3 -2
  13. attn_out_all/layer_21_width_16k_l0_small/config.json +3 -2
  14. attn_out_all/layer_21_width_262k_l0_small/config.json +3 -2
  15. attn_out_all/layer_22_width_16k_l0_small/config.json +3 -2
  16. attn_out_all/layer_22_width_262k_l0_big/config.json +3 -2
  17. attn_out_all/layer_23_width_16k_l0_small/config.json +3 -2
  18. attn_out_all/layer_23_width_262k_l0_big/config.json +3 -2
  19. attn_out_all/layer_23_width_262k_l0_small/config.json +3 -2
  20. attn_out_all/layer_24_width_16k_l0_big/config.json +3 -2
  21. attn_out_all/layer_24_width_262k_l0_big/config.json +3 -2
  22. attn_out_all/layer_25_width_16k_l0_big/config.json +3 -2
  23. attn_out_all/layer_25_width_16k_l0_small/config.json +3 -2
  24. attn_out_all/layer_25_width_262k_l0_small/config.json +3 -2
  25. attn_out_all/layer_26_width_16k_l0_small/config.json +3 -2
  26. attn_out_all/layer_26_width_262k_l0_big/config.json +3 -2
  27. attn_out_all/layer_26_width_262k_l0_small/config.json +3 -2
  28. attn_out_all/layer_27_width_262k_l0_small/config.json +3 -2
  29. attn_out_all/layer_28_width_16k_l0_small/config.json +3 -2
  30. attn_out_all/layer_32_width_262k_l0_small/config.json +3 -2
  31. attn_out_all/layer_33_width_16k_l0_small/config.json +3 -2
  32. attn_out_all/layer_35_width_16k_l0_big/config.json +3 -2
  33. attn_out_all/layer_35_width_262k_l0_small/config.json +3 -2
  34. attn_out_all/layer_36_width_262k_l0_small/config.json +3 -2
  35. attn_out_all/layer_37_width_16k_l0_big/config.json +3 -2
  36. attn_out_all/layer_37_width_262k_l0_big/config.json +3 -2
  37. attn_out_all/layer_37_width_262k_l0_small/config.json +3 -2
  38. attn_out_all/layer_38_width_262k_l0_big/config.json +3 -2
  39. attn_out_all/layer_39_width_16k_l0_small/config.json +3 -2
  40. attn_out_all/layer_3_width_16k_l0_big/config.json +3 -2
  41. attn_out_all/layer_3_width_16k_l0_small/config.json +3 -2
  42. attn_out_all/layer_3_width_262k_l0_big/config.json +3 -2
  43. attn_out_all/layer_40_width_16k_l0_big/config.json +3 -2
  44. attn_out_all/layer_40_width_262k_l0_small/config.json +3 -2
  45. attn_out_all/layer_41_width_262k_l0_big/config.json +3 -2
  46. attn_out_all/layer_43_width_16k_l0_big/config.json +3 -2
  47. attn_out_all/layer_43_width_16k_l0_small/config.json +3 -2
  48. attn_out_all/layer_43_width_262k_l0_small/config.json +3 -2
  49. attn_out_all/layer_44_width_16k_l0_big/config.json +3 -2
  50. attn_out_all/layer_44_width_262k_l0_big/config.json +3 -2
attn_out_all/layer_0_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.0.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.0.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_10_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.10.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.10.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_10_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.10.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.10.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_12_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.12.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.12.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 17,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.12.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.12.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 17,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_13_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.13.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.13.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 18,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.13.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.13.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 18,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_14_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.14.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.14.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 112,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.14.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.14.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 112,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_15_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.15.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.15.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 19,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.15.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.15.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 19,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_16_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_16_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_17_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.17.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.17.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.17.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.17.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_19_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.19.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.19.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.19.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.19.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_1_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 63,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 63,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_21_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.21.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.21.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.21.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.21.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_21_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.21.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.21.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.21.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.21.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_22_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.22.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.22.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.22.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.22.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_22_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.22.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.22.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.22.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.22.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_23_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_23_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_23_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_24_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.24.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.24.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.24.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.24.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_24_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.24.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.24.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.24.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.24.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_25_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_25_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_25_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.25.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.25.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_26_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_26_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_26_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.26.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.26.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_27_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.27.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.27.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.27.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.27.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_28_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.28.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.28.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.28.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.28.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_32_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.32.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.32.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.32.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.32.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_33_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.33.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.33.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.33.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.33.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_35_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.35.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.35.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.35.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.35.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_35_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.35.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.35.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.35.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.35.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_36_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.36.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.36.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.36.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.36.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_37_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_37_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_37_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_38_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.38.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.38.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.38.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.38.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_39_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.39.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.39.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.39.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.39.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_3_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 71,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 71,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_3_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_3_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 71,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 71,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_40_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_40_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_41_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.41.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.41.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.41.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.41.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_43_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_43_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_43_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.43.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.43.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_44_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.44.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.44.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.44.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.44.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_44_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.44.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.44.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.44.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.44.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-12b-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }