Hugging Face's logo

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

SAELens
Model card Files
xet
Community
CallumMcDougallGDM commited on
Commit
6fbcb8e
·
verified ·
1 Parent(s): f702e27

Add files using upload-large-folder tool

Browse files
Files changed (49) hide show
  1. attn_out/layer_12_width_16k_l0_big/config.json +3 -2
  2. attn_out/layer_12_width_16k_l0_medium/config.json +3 -2
  3. attn_out/layer_12_width_16k_l0_small/config.json +3 -2
  4. attn_out/layer_12_width_262k_l0_big/config.json +3 -2
  5. attn_out/layer_12_width_262k_l0_medium/config.json +3 -2
  6. attn_out/layer_12_width_262k_l0_small/config.json +3 -2
  7. attn_out/layer_12_width_65k_l0_big/config.json +3 -2
  8. attn_out/layer_12_width_65k_l0_medium/config.json +3 -2
  9. attn_out/layer_12_width_65k_l0_small/config.json +3 -2
  10. attn_out/layer_15_width_16k_l0_big/config.json +3 -2
  11. attn_out/layer_15_width_16k_l0_medium/config.json +3 -2
  12. attn_out/layer_15_width_16k_l0_small/config.json +3 -2
  13. attn_out/layer_15_width_262k_l0_big/config.json +3 -2
  14. attn_out/layer_15_width_262k_l0_medium/config.json +3 -2
  15. attn_out/layer_15_width_262k_l0_small/config.json +3 -2
  16. attn_out/layer_15_width_65k_l0_big/config.json +3 -2
  17. attn_out/layer_15_width_65k_l0_medium/config.json +3 -2
  18. attn_out/layer_15_width_65k_l0_small/config.json +3 -2
  19. attn_out/layer_5_width_16k_l0_big/config.json +3 -2
  20. attn_out/layer_5_width_16k_l0_medium/config.json +3 -2
  21. attn_out/layer_5_width_16k_l0_small/config.json +3 -2
  22. attn_out/layer_5_width_262k_l0_big/config.json +3 -2
  23. attn_out/layer_5_width_262k_l0_medium/config.json +3 -2
  24. attn_out/layer_5_width_262k_l0_small/config.json +3 -2
  25. attn_out/layer_5_width_65k_l0_big/config.json +3 -2
  26. attn_out/layer_5_width_65k_l0_medium/config.json +3 -2
  27. attn_out/layer_5_width_65k_l0_small/config.json +3 -2
  28. attn_out/layer_9_width_16k_l0_big/config.json +3 -2
  29. attn_out/layer_9_width_16k_l0_medium/config.json +3 -2
  30. attn_out/layer_9_width_16k_l0_small/config.json +3 -2
  31. attn_out/layer_9_width_262k_l0_big/config.json +3 -2
  32. attn_out/layer_9_width_262k_l0_medium/config.json +3 -2
  33. attn_out/layer_9_width_262k_l0_small/config.json +3 -2
  34. attn_out/layer_9_width_65k_l0_big/config.json +3 -2
  35. attn_out/layer_9_width_65k_l0_medium/config.json +3 -2
  36. attn_out/layer_9_width_65k_l0_small/config.json +3 -2
  37. transcoder_all/layer_0_width_262k_l0_big_affine/config.json +3 -2
  38. transcoder_all/layer_10_width_16k_l0_big/config.json +3 -2
  39. transcoder_all/layer_11_width_262k_l0_small/config.json +3 -2
  40. transcoder_all/layer_15_width_262k_l0_big_affine/config.json +3 -2
  41. transcoder_all/layer_17_width_16k_l0_big/config.json +3 -2
  42. transcoder_all/layer_1_width_16k_l0_big_affine/config.json +3 -2
  43. transcoder_all/layer_1_width_262k_l0_small_affine/config.json +3 -2
  44. transcoder_all/layer_2_width_262k_l0_big/config.json +3 -2
  45. transcoder_all/layer_4_width_262k_l0_small_affine/config.json +3 -2
  46. transcoder_all/layer_7_width_16k_l0_small_affine/config.json +3 -2
  47. transcoder_all/layer_9_width_16k_l0_small/config.json +3 -2
  48. transcoder_all/layer_9_width_16k_l0_small_affine/config.json +3 -2
  49. transcoder_all/layer_9_width_262k_l0_small_affine/config.json +3 -2
attn_out/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": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_12_width_16k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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/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": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_12_width_262k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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/layer_12_width_65k_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": 65536,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_12_width_65k_l0_medium/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": 65536,
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.12.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.12.self_attn.o_proj.input",
4
  "width": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_12_width_65k_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": 65536,
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": 65536,
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/layer_15_width_16k_l0_big/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": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_15_width_16k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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/layer_15_width_262k_l0_big/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": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_15_width_262k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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/layer_15_width_65k_l0_big/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": 65536,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_15_width_65k_l0_medium/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": 65536,
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.15.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.15.self_attn.o_proj.input",
4
  "width": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_15_width_65k_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": 65536,
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": 65536,
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/layer_5_width_16k_l0_big/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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 135,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 135,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_5_width_16k_l0_medium/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": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 55,
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": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 55,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_5_width_16k_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": 16384,
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": 16384,
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/layer_5_width_262k_l0_big/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": 135,
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": 135,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_5_width_262k_l0_medium/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": 55,
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": 55,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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/layer_5_width_65k_l0_big/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": 65536,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 135,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 135,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_5_width_65k_l0_medium/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": 65536,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 55,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 55,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_5_width_65k_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": 65536,
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": 65536,
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/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": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_16k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_16k_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": 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.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": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_262k_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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
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": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_262k_l0_medium/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": 60,
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": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/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
  }
attn_out/layer_9_width_65k_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": 65536,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 150,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_65k_l0_medium/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": 65536,
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.9.self_attn.o_proj.input",
3
  "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input",
4
  "width": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 60,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
attn_out/layer_9_width_65k_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": 65536,
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": 65536,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "sae"
10
  }
transcoder_all/layer_0_width_262k_l0_big_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.0.pre_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": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.0.pre_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": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_10_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.10.pre_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.pre_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": "transcoder"
10
  }
transcoder_all/layer_11_width_262k_l0_small/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.11.pre_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.pre_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": "transcoder"
10
  }
transcoder_all/layer_15_width_262k_l0_big_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.15.pre_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": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.15.pre_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": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_17_width_16k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.17.pre_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.pre_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": "transcoder"
10
  }
transcoder_all/layer_1_width_16k_l0_big_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
- "affine_connection": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output",
4
  "width": 16384,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 70,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_1_width_262k_l0_small_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
8
- "affine_connection": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 11,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_2_width_262k_l0_big/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 80,
8
- "affine_connection": false
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 80,
8
+ "affine_connection": false,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_4_width_262k_l0_small_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output",
4
  "width": 262144,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
- "affine_connection": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output",
4
  "width": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 16,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_7_width_16k_l0_small_affine/config.json CHANGED
@@ -2,8 +2,9 @@
2
  "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output",
4
  "width": 16384,
5
- "model_name": "gemma-v3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
- "affine_connection": true
 
9
  }
 
2
  "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output",
3
  "hf_hook_point_out": "model.layers.7.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": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/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-pt",
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-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": false,
9
+ "type": "transcoder"
10
  }
transcoder_all/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-pt",
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-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }
transcoder_all/layer_9_width_262k_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": 262144,
5
- "model_name": "gemma-v3-270m-pt",
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": 262144,
5
+ "model_name": "google/gemma-3-270m-pt",
6
  "architecture": "jump_relu",
7
  "l0": 20,
8
+ "affine_connection": true,
9
+ "type": "transcoder"
10
  }