diff --git a/transcoder/layer_12_width_16k_l0_big/config.json b/transcoder/layer_12_width_16k_l0_big/config.json index 9fb0c11d6f33c338b2fd653c17e46ed2241babed..06e3c6ddb30d5d01f96864a5e462b4a64dc5a263 100644 --- a/transcoder/layer_12_width_16k_l0_big/config.json +++ b/transcoder/layer_12_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_big_affine/config.json b/transcoder/layer_12_width_16k_l0_big_affine/config.json index 92f12e632f96a83f8a6696497cfe4b64f44c25c1..184eb1e64887a5cca9dbdc5d5fa3641b94c84ca0 100644 --- a/transcoder/layer_12_width_16k_l0_big_affine/config.json +++ b/transcoder/layer_12_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_medium/config.json b/transcoder/layer_12_width_16k_l0_medium/config.json index 8d1c5d3c905458361a59e8587f2dff61fbba2a7f..95d1d0c038c5343eeb142ee54f51812e20a550ce 100644 --- a/transcoder/layer_12_width_16k_l0_medium/config.json +++ b/transcoder/layer_12_width_16k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_medium_affine/config.json b/transcoder/layer_12_width_16k_l0_medium_affine/config.json index f9184c27b2715f41ff9919e433626289e84e37ae..1fd8a7fde938228e620338eeb883525df31d0445 100644 --- a/transcoder/layer_12_width_16k_l0_medium_affine/config.json +++ b/transcoder/layer_12_width_16k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_small/config.json b/transcoder/layer_12_width_16k_l0_small/config.json index a5d5c4df5ca8afe94a7090a2da4d49a385c04495..5450100bc84292cbdf0368a004949a63b0e59c8a 100644 --- a/transcoder/layer_12_width_16k_l0_small/config.json +++ b/transcoder/layer_12_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_small_affine/config.json b/transcoder/layer_12_width_16k_l0_small_affine/config.json index 3672cd330403a2c84d868c59ef7c347394018161..67b86e46faa3d05f27106d023cc4748b7c7d3bc1 100644 --- a/transcoder/layer_12_width_16k_l0_small_affine/config.json +++ b/transcoder/layer_12_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_big/config.json b/transcoder/layer_12_width_262k_l0_big/config.json index 57bdede75818fcfc49c8b95cd509e8d42e1fbb78..9f8b0b93581e70138517e2025262fc480a372400 100644 --- a/transcoder/layer_12_width_262k_l0_big/config.json +++ b/transcoder/layer_12_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_big_affine/config.json b/transcoder/layer_12_width_262k_l0_big_affine/config.json index c5a4acc01420f2f888fa352bc3f41bc752f237d9..192a13c3157399b0d8c3c4996ed4b259286bdb80 100644 --- a/transcoder/layer_12_width_262k_l0_big_affine/config.json +++ b/transcoder/layer_12_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_medium/config.json b/transcoder/layer_12_width_262k_l0_medium/config.json index 7feac1e4f0a12d725f9037ba2fb377fcba1caa5a..b7d599fa4494dc5892615c9139edf08d1d06a34f 100644 --- a/transcoder/layer_12_width_262k_l0_medium/config.json +++ b/transcoder/layer_12_width_262k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_medium_affine/config.json b/transcoder/layer_12_width_262k_l0_medium_affine/config.json index 07db6d8f15e7bc0a7c470960f5fcc025481d84f8..ffdf8da059b3b75d9e16dfdcf01782f5f7286638 100644 --- a/transcoder/layer_12_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_12_width_262k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_small/config.json b/transcoder/layer_12_width_262k_l0_small/config.json index ff4b706b0601213dfb6ce631ca8588622d6de333..ab3487fee6b96659eb71ce8a74c041429773974a 100644 --- a/transcoder/layer_12_width_262k_l0_small/config.json +++ b/transcoder/layer_12_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_small_affine/config.json b/transcoder/layer_12_width_262k_l0_small_affine/config.json index 30af5a0539ed1111c9bfea244b62962338155c99..dcbd6ae8309b2e4ef8f017098b92eb8df658705a 100644 --- a/transcoder/layer_12_width_262k_l0_small_affine/config.json +++ b/transcoder/layer_12_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_big/config.json b/transcoder/layer_12_width_65k_l0_big/config.json index a8576ac45b3bc37a7bdbeca092fff36d4f9eb1e1..fa75ccf9b4392a1da0739539d5904eec5c9f548e 100644 --- a/transcoder/layer_12_width_65k_l0_big/config.json +++ b/transcoder/layer_12_width_65k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_big_affine/config.json b/transcoder/layer_12_width_65k_l0_big_affine/config.json index a0a01e70a875da40a6107585cccb727f8a9bd7d5..b48c259c7e47d534f62e46f3e227c7f6e9f7b6d0 100644 --- a/transcoder/layer_12_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_12_width_65k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_medium/config.json b/transcoder/layer_12_width_65k_l0_medium/config.json index 173ffa2b0474c6bd62ad0db417bff25ffaa74eae..9d6702452fc59158f77252aa1d59958bca0cbf03 100644 --- a/transcoder/layer_12_width_65k_l0_medium/config.json +++ b/transcoder/layer_12_width_65k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_medium_affine/config.json b/transcoder/layer_12_width_65k_l0_medium_affine/config.json index 108ee5ce81d13eaec97f4d2f83d0b5797da22575..43a4f5ee710fa7ba604240d8fd3ad39fab5e159a 100644 --- a/transcoder/layer_12_width_65k_l0_medium_affine/config.json +++ b/transcoder/layer_12_width_65k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_small/config.json b/transcoder/layer_12_width_65k_l0_small/config.json index 9254341d47041d5d53b9d52540d388df7f839a64..8145dbce37f9fe0691d6180f64fa320f1e471f82 100644 --- a/transcoder/layer_12_width_65k_l0_small/config.json +++ b/transcoder/layer_12_width_65k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_small_affine/config.json b/transcoder/layer_12_width_65k_l0_small_affine/config.json index 6eac11ff5ed11005e51ea436621d5d3be1b8e0b3..3b5b5204187be1b4ad3a06b3f697bcd87c1e09cc 100644 --- a/transcoder/layer_12_width_65k_l0_small_affine/config.json +++ b/transcoder/layer_12_width_65k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_big/config.json b/transcoder/layer_15_width_16k_l0_big/config.json index 5da520a7b482dc3f08b521db3ff53a7bcaeedf6e..c2329f83ff77fa8355cf5f1c58d69855331c4361 100644 --- a/transcoder/layer_15_width_16k_l0_big/config.json +++ b/transcoder/layer_15_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_big_affine/config.json b/transcoder/layer_15_width_16k_l0_big_affine/config.json index f3e99e59e382d55a92960760fd24b572ffe41ef9..2dee980ab4337185431841fdeff6ad7ce932473c 100644 --- a/transcoder/layer_15_width_16k_l0_big_affine/config.json +++ b/transcoder/layer_15_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_medium/config.json b/transcoder/layer_15_width_16k_l0_medium/config.json index c76eb6f9535cb55bbc303d6047b40691cb1caa7d..3f2b8b845c481968d8cd146d0ea80bfaa90b489f 100644 --- a/transcoder/layer_15_width_16k_l0_medium/config.json +++ b/transcoder/layer_15_width_16k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_medium_affine/config.json b/transcoder/layer_15_width_16k_l0_medium_affine/config.json index 066e5eca348b1919af7ef881cb225dd12a02ea3a..18939e9e64c1bc07dc64d8c6aa111cb2f6ecdfe9 100644 --- a/transcoder/layer_15_width_16k_l0_medium_affine/config.json +++ b/transcoder/layer_15_width_16k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_small/config.json b/transcoder/layer_15_width_16k_l0_small/config.json index 05eb1dae6499b590fda88aeb7e55542b64c8e07d..e2e02ee2754a5eb4afe5cee71e74abbd211a3990 100644 --- a/transcoder/layer_15_width_16k_l0_small/config.json +++ b/transcoder/layer_15_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_small_affine/config.json b/transcoder/layer_15_width_16k_l0_small_affine/config.json index aab3afd6583c80ccd0cc1a339e01ded7751fe0d9..62adde87a777be9d9fe93974f369bbc9e91ef870 100644 --- a/transcoder/layer_15_width_16k_l0_small_affine/config.json +++ b/transcoder/layer_15_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_big/config.json b/transcoder/layer_15_width_262k_l0_big/config.json index 1e22f47bd450b57d3b2b92caa70920b5a5a0d2be..67edad4520ff4989eb7dbbf83c220ddf43503555 100644 --- a/transcoder/layer_15_width_262k_l0_big/config.json +++ b/transcoder/layer_15_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_medium/config.json b/transcoder/layer_15_width_262k_l0_medium/config.json index c25025460e3512f237a2b183445618d6aaec3790..d5e1f18ec74feb49a39732963a955a4f9d07a018 100644 --- a/transcoder/layer_15_width_262k_l0_medium/config.json +++ b/transcoder/layer_15_width_262k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_medium_affine/config.json b/transcoder/layer_15_width_262k_l0_medium_affine/config.json index 6eef705a973c4ac0357f7f695c477b9bab17e6cb..0b8bd39074957147858739070eb15956a935afc5 100644 --- a/transcoder/layer_15_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_15_width_262k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_small/config.json b/transcoder/layer_15_width_262k_l0_small/config.json index 6ca0973e18d53d80d4bfb6e4e80329df9754641e..0ee2239bb3a9be16f29e3b4b0f8463d10266f94d 100644 --- a/transcoder/layer_15_width_262k_l0_small/config.json +++ b/transcoder/layer_15_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_small_affine/config.json b/transcoder/layer_15_width_262k_l0_small_affine/config.json index 629683f948042c8a89727cdc0e9eeafbf919865f..f583fc9e677fa4225a8d9046f157816519bd11fa 100644 --- a/transcoder/layer_15_width_262k_l0_small_affine/config.json +++ b/transcoder/layer_15_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_big/config.json b/transcoder/layer_15_width_65k_l0_big/config.json index 3dbeaeb2d0c671e6727d5c92f9b2e205155d3e4c..d9aabd2ce56b66ba92f3879995d1d89e216d2911 100644 --- a/transcoder/layer_15_width_65k_l0_big/config.json +++ b/transcoder/layer_15_width_65k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_big_affine/config.json b/transcoder/layer_15_width_65k_l0_big_affine/config.json index 6311d31d8b1dd9ad8f892417818ab9b144deb981..08c253b1028cbea72f401afc7ea008dbd296f0f1 100644 --- a/transcoder/layer_15_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_15_width_65k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_medium/config.json b/transcoder/layer_15_width_65k_l0_medium/config.json index 50eb4e82d46c9dc3e0d1d4cc2d74745aa13bb7d6..5feff41e372f89eb6557c85601df65a15f5c2044 100644 --- a/transcoder/layer_15_width_65k_l0_medium/config.json +++ b/transcoder/layer_15_width_65k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_medium_affine/config.json b/transcoder/layer_15_width_65k_l0_medium_affine/config.json index e58ab3cd406d055e3ec9b96b5c346a101e9c6155..06fc12209650dbea7da96f36cd987d8f154c9448 100644 --- a/transcoder/layer_15_width_65k_l0_medium_affine/config.json +++ b/transcoder/layer_15_width_65k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_small/config.json b/transcoder/layer_15_width_65k_l0_small/config.json index c357561f01f44a8f154a9eba7065b96da63b97dd..d9f55153b1f4b5ef863682ad9a2eda4d3acfd7a7 100644 --- a/transcoder/layer_15_width_65k_l0_small/config.json +++ b/transcoder/layer_15_width_65k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_small_affine/config.json b/transcoder/layer_15_width_65k_l0_small_affine/config.json index 12d02ab79428fbd3e6200965cf8d7ce2b48ea3fd..36d75887683fde26ace6d866779b247fa7cfac1f 100644 --- a/transcoder/layer_15_width_65k_l0_small_affine/config.json +++ b/transcoder/layer_15_width_65k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_big/config.json b/transcoder/layer_5_width_16k_l0_big/config.json index 50949faebdcb45dd6de0c7157e84f6601dfa960f..1eaf64df11e004bf88c3d9f57523fa46dad69e41 100644 --- a/transcoder/layer_5_width_16k_l0_big/config.json +++ b/transcoder/layer_5_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_big_affine/config.json b/transcoder/layer_5_width_16k_l0_big_affine/config.json index a7b16594ded7570696ab1962a7bb9ade03902fcd..79f3e2db6501a665b9fbc4346f1f7f7b85208d49 100644 --- a/transcoder/layer_5_width_16k_l0_big_affine/config.json +++ b/transcoder/layer_5_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_medium/config.json b/transcoder/layer_5_width_16k_l0_medium/config.json index daf87612225dbe7f0ea25c6e9813c9bb89a71dbc..89311d5186fbf960744dd37f41b92d85535c73bc 100644 --- a/transcoder/layer_5_width_16k_l0_medium/config.json +++ b/transcoder/layer_5_width_16k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_medium_affine/config.json b/transcoder/layer_5_width_16k_l0_medium_affine/config.json index 2943e58eb6a831bdae4c522feccb9f85d647747c..726f3f286fd082b1ff10ec9f1dfd143497b6d671 100644 --- a/transcoder/layer_5_width_16k_l0_medium_affine/config.json +++ b/transcoder/layer_5_width_16k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_small/config.json b/transcoder/layer_5_width_16k_l0_small/config.json index 4ab893bffbefdd5c03150513515e55d75b6728b2..8caf1abc868b42bc506e3cd16b73185abb4a6ee8 100644 --- a/transcoder/layer_5_width_16k_l0_small/config.json +++ b/transcoder/layer_5_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_small_affine/config.json b/transcoder/layer_5_width_16k_l0_small_affine/config.json index 3b652a2e502f76c3ba30ca603075ceee996ee02d..a04e65893c64dd388e92f140ba873d016059c125 100644 --- a/transcoder/layer_5_width_16k_l0_small_affine/config.json +++ b/transcoder/layer_5_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_big/config.json b/transcoder/layer_5_width_262k_l0_big/config.json index 949bcd5bbd0ecc8f549cb6f742348cf9c18d9f4f..94548d208c98ddc3180ed3c7db0bb05b66edc342 100644 --- a/transcoder/layer_5_width_262k_l0_big/config.json +++ b/transcoder/layer_5_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_big_affine/config.json b/transcoder/layer_5_width_262k_l0_big_affine/config.json index 09f402de6306ff8b6c2ae5974e0897057d794707..ca7b04861cf4604664884147b1ee2ed885b90718 100644 --- a/transcoder/layer_5_width_262k_l0_big_affine/config.json +++ b/transcoder/layer_5_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_medium/config.json b/transcoder/layer_5_width_262k_l0_medium/config.json index a2025a355e8789955d074ea34e8bd8b723fb2ccb..70433a71695d0960db322b046da4fbddca5e7c62 100644 --- a/transcoder/layer_5_width_262k_l0_medium/config.json +++ b/transcoder/layer_5_width_262k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_medium_affine/config.json b/transcoder/layer_5_width_262k_l0_medium_affine/config.json index 74e517298830bb678b25f5caaad80269d049f00a..6ced36b662eb43a654bdb0b0576a522582db6929 100644 --- a/transcoder/layer_5_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_5_width_262k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_small/config.json b/transcoder/layer_5_width_262k_l0_small/config.json index 81925dda7de43d2b0b8f2fd05cbdd7f9a6e9cc9f..5cbb850aabc3ebd0bc2cdaa5d178b8a3767a9266 100644 --- a/transcoder/layer_5_width_262k_l0_small/config.json +++ b/transcoder/layer_5_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_small_affine/config.json b/transcoder/layer_5_width_262k_l0_small_affine/config.json index d45f1ab35b0c4af730fe3a444749450def194010..f0d1fe25ab531ef5773dce810cce5a596c9cb1f8 100644 --- a/transcoder/layer_5_width_262k_l0_small_affine/config.json +++ b/transcoder/layer_5_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_big/config.json b/transcoder/layer_5_width_65k_l0_big/config.json index a102d48ce39ef4090535b15084ec8da99b942701..a491a1de3b9ba8e720efd432ea80ae9072c4695b 100644 --- a/transcoder/layer_5_width_65k_l0_big/config.json +++ b/transcoder/layer_5_width_65k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_big_affine/config.json b/transcoder/layer_5_width_65k_l0_big_affine/config.json index 99d98baa1c44af57a1e31c15811cb26713aab1f8..b5f13a11e2897a519fce9cf4a9a53f53ea3d1185 100644 --- a/transcoder/layer_5_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_5_width_65k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 135, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_medium/config.json b/transcoder/layer_5_width_65k_l0_medium/config.json index 0e2dcf07791e734d647194db0135c78a7bdd062d..ef8fb1c0b72d90f3ef18bafaae997b416ed92e8e 100644 --- a/transcoder/layer_5_width_65k_l0_medium/config.json +++ b/transcoder/layer_5_width_65k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_medium_affine/config.json b/transcoder/layer_5_width_65k_l0_medium_affine/config.json index 8f270e19169d4c743f400249c60ad3d6274a9ae7..6a5582076b09c421bcb3f9cdda1a97b0d08bca40 100644 --- a/transcoder/layer_5_width_65k_l0_medium_affine/config.json +++ b/transcoder/layer_5_width_65k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 55, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_small/config.json b/transcoder/layer_5_width_65k_l0_small/config.json index bbf6d618d5abd53f8ce06172431e9e0642efc39a..ebd27332a39fc2c8fce0931a9b62d225420f13cd 100644 --- a/transcoder/layer_5_width_65k_l0_small/config.json +++ b/transcoder/layer_5_width_65k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_small_affine/config.json b/transcoder/layer_5_width_65k_l0_small_affine/config.json index f6919b790a6f4d952ec451212fcaa3d7de2073cc..2fb66c574735e0818f9c7494d6579d569122af61 100644 --- a/transcoder/layer_5_width_65k_l0_small_affine/config.json +++ b/transcoder/layer_5_width_65k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_big/config.json b/transcoder/layer_9_width_16k_l0_big/config.json index 84004b6b07c07456d4d8ab22f03fe257fa4b22d0..c255c7a57875b6777c7a17250fe67af32bb5d557 100644 --- a/transcoder/layer_9_width_16k_l0_big/config.json +++ b/transcoder/layer_9_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_big_affine/config.json b/transcoder/layer_9_width_16k_l0_big_affine/config.json index bbfc4f6cb0767a8df2eaf82a7a7838a72cbbee87..edf8eb4872f09a9b84924c1c563ec1b42313e3b4 100644 --- a/transcoder/layer_9_width_16k_l0_big_affine/config.json +++ b/transcoder/layer_9_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium/config.json b/transcoder/layer_9_width_16k_l0_medium/config.json index dd9d5f22fb355b060fbae8ffb787e268e7772fea..b7b069bf249d4ee1fa4e681d743ca85e87c2c696 100644 --- a/transcoder/layer_9_width_16k_l0_medium/config.json +++ b/transcoder/layer_9_width_16k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium_affine/config.json b/transcoder/layer_9_width_16k_l0_medium_affine/config.json index 254147376eb0b835c331076860ecc81e794c8f43..707fbf8546518ff220a69706021418a02f0ce822 100644 --- a/transcoder/layer_9_width_16k_l0_medium_affine/config.json +++ b/transcoder/layer_9_width_16k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_small/config.json b/transcoder/layer_9_width_16k_l0_small/config.json index 0622b3dbaf2a8447a84561945acbe96c697f70f8..80a9aa4b7738f7a8b0b59a8384cdd779b3ad15cb 100644 --- a/transcoder/layer_9_width_16k_l0_small/config.json +++ b/transcoder/layer_9_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big/config.json b/transcoder/layer_9_width_262k_l0_big/config.json index a89f943a86b6b5e51c053769ed18360fe68f22d6..e8002fc2f28530dae0fa5de3411307fd50159401 100644 --- a/transcoder/layer_9_width_262k_l0_big/config.json +++ b/transcoder/layer_9_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big_affine/config.json b/transcoder/layer_9_width_262k_l0_big_affine/config.json index 51287538f72b7750cd7c55105a1800c3de615a05..dda2eaffb4d4b2281b7827e5a13b44b880fab76a 100644 --- a/transcoder/layer_9_width_262k_l0_big_affine/config.json +++ b/transcoder/layer_9_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium/config.json b/transcoder/layer_9_width_262k_l0_medium/config.json index a9db20dab2c90149fe73353a78a2b8b0dc3cc7dc..74709b513a8d7bf81f765e59e65212c8d5e33775 100644 --- a/transcoder/layer_9_width_262k_l0_medium/config.json +++ b/transcoder/layer_9_width_262k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium_affine/config.json b/transcoder/layer_9_width_262k_l0_medium_affine/config.json index 64e1ea3bbc65f55f7713d6d68704bc20f691ef98..84309e6cb1d8c30ccbc728fb2a1979afa00c03da 100644 --- a/transcoder/layer_9_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_9_width_262k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small/config.json b/transcoder/layer_9_width_262k_l0_small/config.json index be335e86325cf1ec8440c001c727c504206b2a7b..c9e640a6d1877ae3d5dbb44d797973d752199b00 100644 --- a/transcoder/layer_9_width_262k_l0_small/config.json +++ b/transcoder/layer_9_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small_affine/config.json b/transcoder/layer_9_width_262k_l0_small_affine/config.json index 0b2c450a15f4dd55cc6c5ed9f5dbd5a29d23ab35..b804eb0b54f6e78d10fa40fe1ab3ee2c95df9248 100644 --- a/transcoder/layer_9_width_262k_l0_small_affine/config.json +++ b/transcoder/layer_9_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_big/config.json b/transcoder/layer_9_width_65k_l0_big/config.json index be9b734793e53da6996393017dd26a998ccacd05..4b447dc3038d8be8f554c3ee9c3126bfe4172d2d 100644 --- a/transcoder/layer_9_width_65k_l0_big/config.json +++ b/transcoder/layer_9_width_65k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_big_affine/config.json b/transcoder/layer_9_width_65k_l0_big_affine/config.json index eb5d798b3d75d75f50c2e12dabea839bf9a3b4ee..803b88ca0c93f23657cd15df0185efb29c9f2a49 100644 --- a/transcoder/layer_9_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_9_width_65k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium/config.json b/transcoder/layer_9_width_65k_l0_medium/config.json index 4c554f551f94d5204f50645aae1b74b23905da79..9c324f559a203a008d50687d1d95a129d69b8470 100644 --- a/transcoder/layer_9_width_65k_l0_medium/config.json +++ b/transcoder/layer_9_width_65k_l0_medium/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium_affine/config.json b/transcoder/layer_9_width_65k_l0_medium_affine/config.json index ff0f39ad9127d753ddc7cffc3d1f1987f9eb0e66..ea8daf0eb69381bcc3678133be12c441ee5fa995 100644 --- a/transcoder/layer_9_width_65k_l0_medium_affine/config.json +++ b/transcoder/layer_9_width_65k_l0_medium_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small/config.json b/transcoder/layer_9_width_65k_l0_small/config.json index d1c08ecfc7aed6bfe15044f8c1a4ac866ce5b8db..d9980b5bad0ba847744cc6bc750ac3834bd79dca 100644 --- a/transcoder/layer_9_width_65k_l0_small/config.json +++ b/transcoder/layer_9_width_65k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small_affine/config.json b/transcoder/layer_9_width_65k_l0_small_affine/config.json index f134256c765698733dfe1a9cfad94ae82635ff0d..e46d4baeea598c1500b6f9013a61eaf1c138080a 100644 --- a/transcoder/layer_9_width_65k_l0_small_affine/config.json +++ b/transcoder/layer_9_width_65k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 65536, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file