Hugging Face's logo

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

SAELens
Model card Files
xet
Community
CallumMcDougallGDM commited on
Commit
256fbbe
·
verified ·
1 Parent(s): 3b3c3f0

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_10_width_262k_l0_small/config.json +3 -2
  2. attn_out_all/layer_11_width_262k_l0_big/config.json +3 -2
  3. attn_out_all/layer_12_width_16k_l0_big/config.json +3 -2
  4. attn_out_all/layer_12_width_16k_l0_small/config.json +3 -2
  5. attn_out_all/layer_12_width_262k_l0_big/config.json +3 -2
  6. attn_out_all/layer_12_width_262k_l0_small/config.json +3 -2
  7. attn_out_all/layer_13_width_262k_l0_big/config.json +3 -2
  8. attn_out_all/layer_13_width_262k_l0_small/config.json +3 -2
  9. attn_out_all/layer_14_width_16k_l0_small/config.json +3 -2
  10. attn_out_all/layer_14_width_262k_l0_big/config.json +3 -2
  11. attn_out_all/layer_15_width_16k_l0_small/config.json +3 -2
  12. attn_out_all/layer_15_width_262k_l0_small/config.json +3 -2
  13. attn_out_all/layer_16_width_16k_l0_big/config.json +3 -2
  14. attn_out_all/layer_16_width_16k_l0_small/config.json +3 -2
  15. attn_out_all/layer_16_width_262k_l0_big/config.json +3 -2
  16. attn_out_all/layer_1_width_16k_l0_big/config.json +3 -2
  17. attn_out_all/layer_1_width_16k_l0_small/config.json +3 -2
  18. attn_out_all/layer_1_width_262k_l0_big/config.json +3 -2
  19. attn_out_all/layer_1_width_262k_l0_small/config.json +3 -2
  20. attn_out_all/layer_3_width_262k_l0_big/config.json +3 -2
  21. attn_out_all/layer_4_width_262k_l0_small/config.json +3 -2
  22. attn_out_all/layer_5_width_262k_l0_small/config.json +3 -2
  23. attn_out_all/layer_6_width_16k_l0_big/config.json +3 -2
  24. attn_out_all/layer_6_width_262k_l0_big/config.json +3 -2
  25. attn_out_all/layer_6_width_262k_l0_small/config.json +3 -2
  26. attn_out_all/layer_8_width_16k_l0_small/config.json +3 -2
  27. attn_out_all/layer_9_width_16k_l0_big/config.json +3 -2
  28. attn_out_all/layer_9_width_262k_l0_small/config.json +3 -2
  29. clt/width_262k_l0_big_affine/config.json +10 -0
  30. clt/width_524k_l0_big/config.json +10 -0
  31. clt/width_524k_l0_big_affine/config.json +10 -0
  32. mlp_out_all/layer_0_width_16k_l0_big/config.json +3 -2
  33. mlp_out_all/layer_0_width_16k_l0_small/config.json +3 -2
  34. mlp_out_all/layer_0_width_262k_l0_big/config.json +3 -2
  35. mlp_out_all/layer_0_width_262k_l0_small/config.json +3 -2
  36. mlp_out_all/layer_10_width_16k_l0_big/config.json +3 -2
  37. mlp_out_all/layer_10_width_16k_l0_small/config.json +3 -2
  38. mlp_out_all/layer_10_width_262k_l0_big/config.json +3 -2
  39. mlp_out_all/layer_11_width_16k_l0_big/config.json +3 -2
  40. mlp_out_all/layer_11_width_16k_l0_small/config.json +3 -2
  41. mlp_out_all/layer_11_width_262k_l0_small/config.json +3 -2
  42. mlp_out_all/layer_13_width_16k_l0_big/config.json +3 -2
  43. mlp_out_all/layer_13_width_16k_l0_small/config.json +3 -2
  44. mlp_out_all/layer_14_width_16k_l0_big/config.json +3 -2
  45. mlp_out_all/layer_14_width_262k_l0_small/config.json +3 -2
  46. mlp_out_all/layer_15_width_16k_l0_big/config.json +3 -2
  47. mlp_out_all/layer_15_width_262k_l0_big/config.json +3 -2
  48. mlp_out_all/layer_16_width_262k_l0_small/config.json +3 -2
  49. mlp_out_all/layer_17_width_16k_l0_big/config.json +3 -2
  50. mlp_out_all/layer_17_width_16k_l0_small/config.json +3 -2
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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_11_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.11.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.11.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.11.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_12_width_16k_l0_big/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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_12_width_16k_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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_12_width_262k_l0_big/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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_13_width_262k_l0_big/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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_13_width_262k_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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_14_width_16k_l0_small/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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_14_width_262k_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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_15_width_262k_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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
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": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_16_width_16k_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": 16384,
5
- "model_name": "gemma-v3-270m-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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
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-v3-270m-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-3-270m-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-v3-270m-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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_1_width_16k_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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_1_width_16k_l0_small/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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_1_width_262k_l0_small/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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
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-3-270m-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-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 90,
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-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 90,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_4_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.4.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.4.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.4.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.4.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_5_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.5.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.5.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 18,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.5.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.5.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 18,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_6_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_6_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_6_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_8_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.8.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.8.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.8.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.8.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_9_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out_all/layer_9_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
clt/width_262k_l0_big_affine/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output",
3
+ "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output",
4
+ "width": 262080,
5
+ "model_name": "google/gemma-3-270m-pt",
6
+ "architecture": "jump_relu",
7
+ "l0": 150,
8
+ "affine_connection": true,
9
+ "type": "clt"
10
+ }
clt/width_524k_l0_big/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output",
3
+ "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output",
4
+ "width": 524160,
5
+ "model_name": "google/gemma-3-270m-pt",
6
+ "architecture": "jump_relu",
7
+ "l0": 150,
8
+ "affine_connection": false,
9
+ "type": "clt"
10
+ }
clt/width_524k_l0_big_affine/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output",
3
+ "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output",
4
+ "width": 524160,
5
+ "model_name": "google/gemma-3-270m-pt",
6
+ "architecture": "jump_relu",
7
+ "l0": 150,
8
+ "affine_connection": true,
9
+ "type": "clt"
10
+ }
mlp_out_all/layer_0_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_0_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_0_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_0_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 10,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_10_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_10_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_10_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_11_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_11_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_11_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_13_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_13_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_14_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_14_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_15_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_15_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_16_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_17_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 120,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
mlp_out_all/layer_17_width_16k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }