Instructions to use google/gemma-scope-2-270m-pt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- SAELens
How to use google/gemma-scope-2-270m-pt with SAELens:
# pip install sae-lens from sae_lens import SAE sae, cfg_dict, sparsity = SAE.from_pretrained( release = "RELEASE_ID", # e.g., "gpt2-small-res-jb". See other options in https://github.com/jbloomAus/SAELens/blob/main/sae_lens/pretrained_saes.yaml sae_id = "SAE_ID", # e.g., "blocks.8.hook_resid_pre". Won't always be a hook point ) - Notebooks
- Google Colab
- Kaggle
Add files using upload-large-folder tool
Browse filesThis view is limited to 50 files because it contains too many changes. See raw diff
- crosscoder/layer_5_9_12_15_width_1m_l0_big/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_1m_l0_medium/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_262k_l0_big/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_262k_l0_medium/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_524k_l0_big/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_524k_l0_medium/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_65k_l0_big/config.json +5 -4
- crosscoder/layer_5_9_12_15_width_65k_l0_medium/config.json +5 -4
- mlp_out/layer_12_width_16k_l0_big/config.json +3 -2
- mlp_out/layer_12_width_16k_l0_medium/config.json +3 -2
- mlp_out/layer_12_width_16k_l0_small/config.json +3 -2
- mlp_out/layer_12_width_262k_l0_big/config.json +3 -2
- mlp_out/layer_12_width_262k_l0_medium/config.json +3 -2
- mlp_out/layer_12_width_262k_l0_small/config.json +3 -2
- mlp_out/layer_12_width_65k_l0_big/config.json +3 -2
- mlp_out/layer_12_width_65k_l0_medium/config.json +3 -2
- mlp_out/layer_12_width_65k_l0_small/config.json +3 -2
- mlp_out/layer_15_width_16k_l0_big/config.json +3 -2
- mlp_out/layer_15_width_16k_l0_medium/config.json +3 -2
- mlp_out/layer_15_width_16k_l0_small/config.json +3 -2
- mlp_out/layer_15_width_262k_l0_big/config.json +3 -2
- mlp_out/layer_15_width_262k_l0_medium/config.json +3 -2
- mlp_out/layer_15_width_262k_l0_small/config.json +3 -2
- mlp_out/layer_15_width_65k_l0_big/config.json +3 -2
- mlp_out/layer_15_width_65k_l0_medium/config.json +3 -2
- mlp_out/layer_15_width_65k_l0_small/config.json +3 -2
- mlp_out/layer_5_width_16k_l0_big/config.json +3 -2
- mlp_out/layer_5_width_16k_l0_medium/config.json +3 -2
- mlp_out/layer_5_width_16k_l0_small/config.json +3 -2
- mlp_out/layer_5_width_262k_l0_big/config.json +3 -2
- mlp_out/layer_5_width_262k_l0_medium/config.json +3 -2
- mlp_out/layer_5_width_262k_l0_small/config.json +3 -2
- mlp_out/layer_5_width_65k_l0_big/config.json +3 -2
- mlp_out/layer_5_width_65k_l0_medium/config.json +3 -2
- mlp_out/layer_5_width_65k_l0_small/config.json +3 -2
- mlp_out/layer_9_width_16k_l0_big/config.json +3 -2
- mlp_out/layer_9_width_16k_l0_medium/config.json +3 -2
- mlp_out/layer_9_width_16k_l0_small/config.json +3 -2
- mlp_out/layer_9_width_262k_l0_big/config.json +3 -2
- mlp_out/layer_9_width_262k_l0_medium/config.json +3 -2
- mlp_out/layer_9_width_262k_l0_small/config.json +3 -2
- mlp_out/layer_9_width_65k_l0_big/config.json +3 -2
- mlp_out/layer_9_width_65k_l0_medium/config.json +3 -2
- mlp_out/layer_9_width_65k_l0_small/config.json +3 -2
- mlp_out_all/layer_10_width_262k_l0_small/config.json +3 -2
- mlp_out_all/layer_11_width_262k_l0_big/config.json +3 -2
- mlp_out_all/layer_12_width_16k_l0_big/config.json +3 -2
- mlp_out_all/layer_12_width_16k_l0_small/config.json +3 -2
- mlp_out_all/layer_12_width_262k_l0_big/config.json +3 -2
- mlp_out_all/layer_12_width_262k_l0_small/config.json +3 -2
crosscoder/layer_5_9_12_15_width_1m_l0_big/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 1048576,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 1048576,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_1m_l0_medium/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 1048576,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 1048576,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_262k_l0_big/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 262144,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_262k_l0_medium/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 262144,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_524k_l0_big/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 524288,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 524288,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_524k_l0_medium/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 524288,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 524288,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_65k_l0_big/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 65536,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
crosscoder/layer_5_9_12_15_width_65k_l0_medium/config.json
CHANGED
|
@@ -1,9 +1,10 @@
|
|
| 1 |
{
|
| 2 |
-
"hf_hook_point_in": "model.layers.{5
|
| 3 |
-
"hf_hook_point_out": "model.layers.{5
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"hf_hook_point_in": "model.layers.{5,9,12,15}.output",
|
| 3 |
+
"hf_hook_point_out": "model.layers.{5,9,12,15}.output",
|
| 4 |
"width": 65536,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 50,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "crosscoder"
|
| 10 |
}
|
mlp_out/layer_12_width_16k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_12_width_16k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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/layer_12_width_16k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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/layer_12_width_262k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_12_width_262k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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/layer_12_width_262k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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/layer_12_width_65k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_12_width_65k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_12_width_65k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 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": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_16k_l0_medium/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 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": 60,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_16k_l0_small/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 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": 20,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 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": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_262k_l0_medium/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 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": 60,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_262k_l0_small/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-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 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": 20,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_65k_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": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 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": 65536,
|
| 5 |
+
"model_name": "google/gemma-3-270m-pt",
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_15_width_65k_l0_medium/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": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 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": 65536,
|
| 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/layer_15_width_65k_l0_small/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": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 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": 65536,
|
| 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/layer_5_width_16k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 135,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_16k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 55,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_16k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 18,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_262k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 135,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_262k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 55,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_262k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 18,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_65k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 135,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_65k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 55,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_5_width_65k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 18,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_9_width_16k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_16k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": 60,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_16k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_262k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": 150,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_262k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": 60,
|
| 8 |
+
"affine_connection": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_262k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_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": false,
|
| 9 |
+
"type": "sae"
|
| 10 |
}
|
mlp_out/layer_9_width_65k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 150,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_9_width_65k_l0_medium/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 60,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out/layer_9_width_65k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 4 |
"width": 65536,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output",
|
| 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 |
}
|
mlp_out_all/layer_10_width_262k_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": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 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": 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_11_width_262k_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": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 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": 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_12_width_16k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 120,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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_12_width_16k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 16384,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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_12_width_262k_l0_big/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 120,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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_12_width_262k_l0_small/config.json
CHANGED
|
@@ -2,8 +2,9 @@
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output",
|
| 4 |
"width": 262144,
|
| 5 |
-
"model_name": "gemma-
|
| 6 |
"architecture": "jump_relu",
|
| 7 |
"l0": 20,
|
| 8 |
-
"affine_connection": false
|
|
|
|
| 9 |
}
|
|
|
|
| 2 |
"hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output",
|
| 3 |
"hf_hook_point_out": "model.layers.12.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 |
}
|