diff --git a/transcoder_all/layer_0_width_262k_l0_small_affine/config.json b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json index 6b1f1f961edaac20724af406a5765dc7c830da4f..4ce5671171b532efc004cfff774ce06aca38c864 100644 --- a/transcoder_all/layer_0_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 10, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small/config.json b/transcoder_all/layer_10_width_16k_l0_small/config.json index d6d3a065df60dadde5e00f47d825f2dcd2c07640..58e9aec27f7195dc1d3a3e5f56140a4918718ffb 100644 --- a/transcoder_all/layer_10_width_16k_l0_small/config.json +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 16, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small_affine/config.json b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json index cddf61a45ea04e644b2b45542b6c00f0bbfff1c9..0cae63c48eba11dd1e3e49725a9658a67b24d387 100644 --- a/transcoder_all/layer_10_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 16, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_big/config.json b/transcoder_all/layer_10_width_262k_l0_big/config.json index 01e3bfd6837418d6f2fa40a490e4a44db341209d..b3bbdc3ab3f8889ed730f061aa399c48166af788 100644 --- a/transcoder_all/layer_10_width_262k_l0_big/config.json +++ b/transcoder_all/layer_10_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 97, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big/config.json b/transcoder_all/layer_11_width_16k_l0_big/config.json index c5c2998f6fbfc3adf0a166f24449953d03e26d47..cccafcc5de5538bb19ed1fc30e4f0b5a2988ccce 100644 --- a/transcoder_all/layer_11_width_16k_l0_big/config.json +++ b/transcoder_all/layer_11_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 101, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json index a1f51878a06389daac806302375e519e97fff0a2..6388a27d2bfadea0834ce2836de98061395be81e 100644 --- a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 101, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_small_affine/config.json b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json index 5ade9f5d54f26d854792eb2de2ce2d82d02b9cec..652f3d3f5be916c132762ded844df7892aee58a4 100644 --- a/transcoder_all/layer_11_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 16, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_big/config.json b/transcoder_all/layer_12_width_262k_l0_big/config.json index ff66566ae368692f14a745a6cfce1a2b1bd7e8bc..080ee82b7eab715a4f7de3ba91977100eb744e42 100644 --- a/transcoder_all/layer_12_width_262k_l0_big/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 105, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_small/config.json b/transcoder_all/layer_12_width_262k_l0_small/config.json index d419a3dec39538ea772698808be9b8d73e9fe7fa..8a21db0cc56a53b958dcd216c74aea0632207181 100644 --- a/transcoder_all/layer_12_width_262k_l0_small/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 17, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_small_affine/config.json b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json index ebaff3b41a04ead4b76c4b377e6e89147753035c..0b60b2bb8a45a2a01fe3b0825b84f011235bb58e 100644 --- a/transcoder_all/layer_13_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_big/config.json b/transcoder_all/layer_13_width_262k_l0_big/config.json index 6e30c15323ce16bf1287ea779d2814db68837b06..c3dda297354cc366dd6da508923a14cd994d528a 100644 --- a/transcoder_all/layer_13_width_262k_l0_big/config.json +++ b/transcoder_all/layer_13_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 108, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big_affine/config.json b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json index 10c8c0ede273acdb575b5171e3c37b881712e7a6..4cd02e35cd56650801cc71d7901ef471c0cdf28d 100644 --- a/transcoder_all/layer_14_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 112, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_small_affine/config.json b/transcoder_all/layer_14_width_16k_l0_small_affine/config.json index 5419f4514af67e5042dfbd360ae36b101b1ca71f..d8d468d078591c8ead004a07ae3bdc8f131d0e2a 100644 --- a/transcoder_all/layer_14_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_14_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_small/config.json b/transcoder_all/layer_15_width_16k_l0_small/config.json index d6a4714c2c17622d5f287da104d6d3e01244a212..4d4e6a7750328a6ea1084f1b8df6f9d7f35a1fdc 100644 --- a/transcoder_all/layer_15_width_16k_l0_small/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 19, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_small_affine/config.json b/transcoder_all/layer_15_width_262k_l0_small_affine/config.json index c151cf01f236ee0038a3bae271b3e72aaef10373..73918ab5b2b1e49f2ca39f4e9a38e0a8415d7b8d 100644 --- a/transcoder_all/layer_15_width_262k_l0_small_affine/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 19, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big/config.json b/transcoder_all/layer_16_width_262k_l0_big/config.json index cfda1579f851d9d23bc4db6ae9b464c36ec4aa28..96c6cdcd5c216442d3390faf209313268a35f0ec 100644 --- a/transcoder_all/layer_16_width_262k_l0_big/config.json +++ b/transcoder_all/layer_16_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json index 5b9468fdcdee2aa6689dd71faa39762dfd7e8aa3..3acef6cdf2f5531681913af471288bb31cd33dff 100644 --- a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_big_affine/config.json b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json index 91dc5f2da5040a098260ae791e561166c22f0557..01628349205b9bb21c4e59956d1a4ed1e201eb24 100644 --- a/transcoder_all/layer_17_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_18_width_16k_l0_big_affine/config.json b/transcoder_all/layer_18_width_16k_l0_big_affine/config.json index 356f6d172ce4544dd0d4baed22ac724bd5f2790c..05288735482ea1f9d69c7671fff0e782dea6dd0f 100644 --- a/transcoder_all/layer_18_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_18_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big/config.json b/transcoder_all/layer_18_width_262k_l0_big/config.json index c0319077a44c401d4f0680806239c0c2586f7795..09c3b8de366d9a134330d6d38a9adf997fecee5e 100644 --- a/transcoder_all/layer_18_width_262k_l0_big/config.json +++ b/transcoder_all/layer_18_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json index d26699d2da53b30a9b896b9bdd3ab459238acf91..13de15a74bd23dbfdc3e90b82404b9ed53e1a624 100644 --- a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_small/config.json b/transcoder_all/layer_18_width_262k_l0_small/config.json index b16e900fded88363ee5d14e3815790b7db9983ed..7dbad0fd621785b1045985a6d7e4a5dd892a535c 100644 --- a/transcoder_all/layer_18_width_262k_l0_small/config.json +++ b/transcoder_all/layer_18_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_big/config.json b/transcoder_all/layer_19_width_16k_l0_big/config.json index 70d32cda57bf3d7039f07f06e178e36f72791444..5cb4011850908d3377081079cc7496f3c2b594ea 100644 --- a/transcoder_all/layer_19_width_16k_l0_big/config.json +++ b/transcoder_all/layer_19_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_big_affine/config.json b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json index ba804e1a02cc3c41450fabbc2f81a2e540fcb3dd..c5d8349bf3c928d7f625ec5adb383dc0c0ed0bda 100644 --- a/transcoder_all/layer_19_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_small_affine/config.json b/transcoder_all/layer_19_width_16k_l0_small_affine/config.json index 8d176fba0b210f16eada272a11d490e4c7d1fca4..8d31e36143311f681d5679b862daef536781bb5b 100644 --- a/transcoder_all/layer_19_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_19_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_19_width_262k_l0_small_affine/config.json b/transcoder_all/layer_19_width_262k_l0_small_affine/config.json index cd21acca312f8f175231f9db6a37f8441aa7e646..5ecad54d159e14a3b0689ea7690fbea850ca0782 100644 --- a/transcoder_all/layer_19_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_19_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_big_affine/config.json b/transcoder_all/layer_1_width_262k_l0_big_affine/config.json index e8c69b627c5a4c046fd88d9ad59e8fbbb8f00372..046cd8239c2b5d1b5e838322912a69138eb246f2 100644 --- a/transcoder_all/layer_1_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_1_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 63, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_big_affine/config.json b/transcoder_all/layer_20_width_16k_l0_big_affine/config.json index ecbc0b9cfed36526f9f691c483417763c1ab9115..b7aacff4cb50500ac72ba1934974f15de125a1a6 100644 --- a/transcoder_all/layer_20_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_20_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_small/config.json b/transcoder_all/layer_20_width_16k_l0_small/config.json index d911e855212aefec59d3224e4303b3aafb7cfcec..3e55e73dd0fed4101717f92785aa6f535c0e9c8f 100644 --- a/transcoder_all/layer_20_width_16k_l0_small/config.json +++ b/transcoder_all/layer_20_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_small_affine/config.json b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json index ac7374f81729c980cc25177e6c8235ec5b403ecd..3f3bfdf70d1c0b2cd5d529150c7f9ccd717c32f6 100644 --- a/transcoder_all/layer_20_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_big_affine/config.json b/transcoder_all/layer_21_width_16k_l0_big_affine/config.json index 4de5908e9c7436c03771e0f8d4aa359f3cb8382d..efd1174cf4197b5a433493424908dcb2d1281ae0 100644 --- a/transcoder_all/layer_21_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_21_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_small/config.json b/transcoder_all/layer_21_width_16k_l0_small/config.json index 0653660594fafecc8d1132a5e8c6d6eddd13b539..e1cc24509229ae2eb919654da1e1c80667ed29ca 100644 --- a/transcoder_all/layer_21_width_16k_l0_small/config.json +++ b/transcoder_all/layer_21_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_small/config.json b/transcoder_all/layer_22_width_16k_l0_small/config.json index 5713d5e517fdf058aeb4c73c696562338d05c7e5..832bcdb2cc163fb521ae4b5023300c001edb79e8 100644 --- a/transcoder_all/layer_22_width_16k_l0_small/config.json +++ b/transcoder_all/layer_22_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_big_affine/config.json b/transcoder_all/layer_24_width_16k_l0_big_affine/config.json index ff51a972ac57d0b5793d732cc06bef8c50a360b9..64bca718fa584e478d5b4b27db6d1534987aa1b9 100644 --- a/transcoder_all/layer_24_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_24_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_small/config.json b/transcoder_all/layer_24_width_16k_l0_small/config.json index 8f6355d944cae1e1f9273145041e2067287ab730..432872ca6144086e2a54a557c6cf88bfa153130b 100644 --- a/transcoder_all/layer_24_width_16k_l0_small/config.json +++ b/transcoder_all/layer_24_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_big_affine/config.json b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json index 28ac02657a51a8af5f0d9024ce5f83ea7abf5397..9f34b5fa093c8ba4fa59872238115e87ac7b6a5f 100644 --- a/transcoder_all/layer_24_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_small/config.json b/transcoder_all/layer_24_width_262k_l0_small/config.json index 4668dbaad6cde1d489fb524d9f39f008de3fc2b4..e8bb618f26a754872872b66d1359ec22c98d7c30 100644 --- a/transcoder_all/layer_24_width_262k_l0_small/config.json +++ b/transcoder_all/layer_24_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_small_affine/config.json b/transcoder_all/layer_24_width_262k_l0_small_affine/config.json index e26de7cdf008f2943d91e0ad6ef03e89a7d3abee..fc33f3cf0c9526a867475cd0115433ce7b3a71ee 100644 --- a/transcoder_all/layer_24_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_24_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_big/config.json b/transcoder_all/layer_25_width_16k_l0_big/config.json index 5920b1cf258758d5751ae2db7a43c246bc4ad2b4..00a95a5c252ea127a14e7c145e2b6b3ee23ba86d 100644 --- a/transcoder_all/layer_25_width_16k_l0_big/config.json +++ b/transcoder_all/layer_25_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_small/config.json b/transcoder_all/layer_25_width_16k_l0_small/config.json index 5640e423a58a44b3915a55bc55dc87c64948f15b..b45993a75381ea1c83922246604cb196a60747d1 100644 --- a/transcoder_all/layer_25_width_16k_l0_small/config.json +++ b/transcoder_all/layer_25_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_27_width_16k_l0_small/config.json b/transcoder_all/layer_27_width_16k_l0_small/config.json index 4cb4ea551380004a6e49b3d17057ad2ed70f4bd2..0d6ca01efaef632b3c2d9994045fa35abe73f5c8 100644 --- a/transcoder_all/layer_27_width_16k_l0_small/config.json +++ b/transcoder_all/layer_27_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_28_width_16k_l0_small/config.json b/transcoder_all/layer_28_width_16k_l0_small/config.json index 8a592f0c6633d741bca7a92c302cb2aa5b36b684..9c9c8e1e78726a21d20cf89449c6e3f9afc12e39 100644 --- a/transcoder_all/layer_28_width_16k_l0_small/config.json +++ b/transcoder_all/layer_28_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.28.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.28.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_small_affine/config.json b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json index c32326bfa2a58a09a0c6c3a84e4453699388be19..df3bcf0c460927c56720682a546317f30afceba7 100644 --- a/transcoder_all/layer_29_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_big/config.json b/transcoder_all/layer_30_width_16k_l0_big/config.json index 5b7bb4beea5627a5fbe8baa182b03551678d8774..4dcebc400e8fa42f948128432feb895e426f7e0b 100644 --- a/transcoder_all/layer_30_width_16k_l0_big/config.json +++ b/transcoder_all/layer_30_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_big_affine/config.json b/transcoder_all/layer_30_width_16k_l0_big_affine/config.json index b8a70aab4c2f7b171cba913b5974e10e0f869c4b..d4b511e567f6ba1650b2dde5744b22d29981ebaa 100644 --- a/transcoder_all/layer_30_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_30_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_small/config.json b/transcoder_all/layer_30_width_16k_l0_small/config.json index cb89d901882f4e1ca9169d96ed00379eba67f621..f7a53dd169b17c4ef520c6f7119537231ba9f5ed 100644 --- a/transcoder_all/layer_30_width_16k_l0_small/config.json +++ b/transcoder_all/layer_30_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_small_affine/config.json b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json index 81d583396d7503f580fe1acdcedf5527d7640369..f4f45b5b07fd01d246113e797a21342b804172ba 100644 --- a/transcoder_all/layer_30_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_big_affine/config.json b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json index 438fd33bc60c32472577e1b56530c1f6d237b91f..abec5cde658f111658a3ec04576c275b8a02fa44 100644 --- a/transcoder_all/layer_31_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_31_width_262k_l0_big/config.json b/transcoder_all/layer_31_width_262k_l0_big/config.json index 5231232add1e3aa7e5244e84ab16c6192ae96cb8..90d9b49255909e55824a24769c3995f2ec4e5f71 100644 --- a/transcoder_all/layer_31_width_262k_l0_big/config.json +++ b/transcoder_all/layer_31_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_big/config.json b/transcoder_all/layer_32_width_16k_l0_big/config.json index 5aa3a4a6b282123b7f3a0d63544b70a77d0cf8cd..24f426199e7a981306ddcdd631fec5976c779615 100644 --- a/transcoder_all/layer_32_width_16k_l0_big/config.json +++ b/transcoder_all/layer_32_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_small/config.json b/transcoder_all/layer_32_width_16k_l0_small/config.json index 265b71e9e24c693d80c52707c32aa82953b8abe8..59dae02297b25e46a847042e7140d72e6e270fd3 100644 --- a/transcoder_all/layer_32_width_16k_l0_small/config.json +++ b/transcoder_all/layer_32_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_big_affine/config.json b/transcoder_all/layer_32_width_262k_l0_big_affine/config.json index 988f1460650269638a75d399a1032d3ff69c7c35..120ca9de53dd27a4e9cf57bc1feaf08659a6bf55 100644 --- a/transcoder_all/layer_32_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_32_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_small_affine/config.json b/transcoder_all/layer_32_width_262k_l0_small_affine/config.json index 9c5b762b6988046d1295284d8c7ee6fa2ce250f4..fd5a6dabe0c682a5852e60a6392c196e82482eb3 100644 --- a/transcoder_all/layer_32_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_32_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_33_width_16k_l0_big/config.json b/transcoder_all/layer_33_width_16k_l0_big/config.json index 34cbf2b74ccd7d59a4858aa9785f2439333c92a2..df65ffd1347927bf0ff09c73f507386b9c80c342 100644 --- a/transcoder_all/layer_33_width_16k_l0_big/config.json +++ b/transcoder_all/layer_33_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_small/config.json b/transcoder_all/layer_33_width_262k_l0_small/config.json index bb31be42fab7ee1190859d147c582272760dfc7b..83752d8c7ce47f74d96985606c3b6f1579c9b365 100644 --- a/transcoder_all/layer_33_width_262k_l0_small/config.json +++ b/transcoder_all/layer_33_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_34_width_16k_l0_big_affine/config.json b/transcoder_all/layer_34_width_16k_l0_big_affine/config.json index 809c1fb73cb791dd9e546380da99378a2c35d3b5..beb2df7de4c4555405db1d208c0982f67051a124 100644 --- a/transcoder_all/layer_34_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_34_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_35_width_16k_l0_big_affine/config.json b/transcoder_all/layer_35_width_16k_l0_big_affine/config.json index 5d0976fb1afd74915a0716e47b1c1a602b23be35..3423c9b3cee7b5b5117a542a1a01eb291ad9d2e0 100644 --- a/transcoder_all/layer_35_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_35_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.35.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.35.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_35_width_262k_l0_small/config.json b/transcoder_all/layer_35_width_262k_l0_small/config.json index 5205cb9367df3a05b07503a2e6b78b0216df99df..e0be3f1e3ad01d21a74d2a790405434e76b54da4 100644 --- a/transcoder_all/layer_35_width_262k_l0_small/config.json +++ b/transcoder_all/layer_35_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.35.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.35.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_36_width_262k_l0_big/config.json b/transcoder_all/layer_36_width_262k_l0_big/config.json index f9a77dffe529f00fee3df7f3322366eee0fd4290..07847c8625197de58e3012ad501caae17d84cbd5 100644 --- a/transcoder_all/layer_36_width_262k_l0_big/config.json +++ b/transcoder_all/layer_36_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_small/config.json b/transcoder_all/layer_37_width_16k_l0_small/config.json index e36e9117110c842222dd3a169ab5b6851767b353..7ec9df93e8f2a1389ee1249d456047340d707d71 100644 --- a/transcoder_all/layer_37_width_16k_l0_small/config.json +++ b/transcoder_all/layer_37_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_37_width_262k_l0_small/config.json b/transcoder_all/layer_37_width_262k_l0_small/config.json index e598f65f4c8ad8802fd28e9f2d650f6f16359b25..e7742aef5bbfa4f3429a08bcec325e2960668093 100644 --- a/transcoder_all/layer_37_width_262k_l0_small/config.json +++ b/transcoder_all/layer_37_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_38_width_16k_l0_small/config.json b/transcoder_all/layer_38_width_16k_l0_small/config.json index 8e9189ded970a7d2e43e8b16a0c040977a318d19..9d9fb5b8fede2de60527f22abd84bf4cbc223480 100644 --- a/transcoder_all/layer_38_width_16k_l0_small/config.json +++ b/transcoder_all/layer_38_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.38.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_39_width_16k_l0_big/config.json b/transcoder_all/layer_39_width_16k_l0_big/config.json index cabe1abdcddc193888918fc6db6b6f350c27fe90..5d259af73434adae602fb801599eb125be3d4771 100644 --- a/transcoder_all/layer_39_width_16k_l0_big/config.json +++ b/transcoder_all/layer_39_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_39_width_262k_l0_big/config.json b/transcoder_all/layer_39_width_262k_l0_big/config.json index af15f28b0e51fc88b58701715b94ac22c1192e1a..db31d1db8d57054c2a4d886d825a7994ea0f33f8 100644 --- a/transcoder_all/layer_39_width_262k_l0_big/config.json +++ b/transcoder_all/layer_39_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_39_width_262k_l0_small/config.json b/transcoder_all/layer_39_width_262k_l0_small/config.json index 7c01f39fe3d1d04728b412a873a8ca9880553dba..77da660b999ff1f3a7dff9b91d4fb1291bc7c158 100644 --- a/transcoder_all/layer_39_width_262k_l0_small/config.json +++ b/transcoder_all/layer_39_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_big_affine/config.json b/transcoder_all/layer_3_width_16k_l0_big_affine/config.json index 40cb025bfdf3811dcbf388e148c5b2ec91b2a66f..92d54c615575f0c1a7b026050196f3b15f683aa0 100644 --- a/transcoder_all/layer_3_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_3_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 71, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_big/config.json b/transcoder_all/layer_3_width_262k_l0_big/config.json index cdea2e43f7f63a91251c322e124bcdcd64cac5fd..1559f84e9cb469fe9a9605cbb9429db770630938 100644 --- a/transcoder_all/layer_3_width_262k_l0_big/config.json +++ b/transcoder_all/layer_3_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 71, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_big_affine/config.json b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json index 71ff9b78b8e26dc3a8efcab477dc5b1f711f340e..2f0475f513781e77d2eac0208b3233b3bf7e9005 100644 --- a/transcoder_all/layer_3_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 71, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small/config.json b/transcoder_all/layer_3_width_262k_l0_small/config.json index 0a77bbfc6f8f5a68682dda9a79404efb0220222f..f1a5f7ccc5b03f94ed2d885fe3183cfe03329586 100644 --- a/transcoder_all/layer_3_width_262k_l0_small/config.json +++ b/transcoder_all/layer_3_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 11, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_40_width_16k_l0_big/config.json b/transcoder_all/layer_40_width_16k_l0_big/config.json index ea29e847d759daba093ca7156b7b24d6789d1d97..5848c6cbae83cbc67b196186e1e4b5ce2c4e0d88 100644 --- a/transcoder_all/layer_40_width_16k_l0_big/config.json +++ b/transcoder_all/layer_40_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_40_width_16k_l0_small/config.json b/transcoder_all/layer_40_width_16k_l0_small/config.json index 3f9580891534973ef35abe9fd414978da7ef91a6..6d9d533688672fdad377a1bd283606063d6057cd 100644 --- a/transcoder_all/layer_40_width_16k_l0_small/config.json +++ b/transcoder_all/layer_40_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_40_width_262k_l0_big_affine/config.json b/transcoder_all/layer_40_width_262k_l0_big_affine/config.json index 98809ee53e1629cc6239cb0fbad53543f3b88b19..d9354e81779853abcbc89a78483ef751f94566c6 100644 --- a/transcoder_all/layer_40_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_40_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_41_width_16k_l0_big_affine/config.json b/transcoder_all/layer_41_width_16k_l0_big_affine/config.json index d46b77fb5253f885007b644665f3d44fe2fd04d0..0ab565497c2f85f60f6da53fcb06390540dffed5 100644 --- a/transcoder_all/layer_41_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_41_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_41_width_262k_l0_small_affine/config.json b/transcoder_all/layer_41_width_262k_l0_small_affine/config.json index 1cb3f51dc807107027373a7ad6d3edba8c46c9b2..bc894fc5a0c90888e2cf44fc1876530287c96432 100644 --- a/transcoder_all/layer_41_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_41_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_42_width_16k_l0_big_affine/config.json b/transcoder_all/layer_42_width_16k_l0_big_affine/config.json index 692906fe556283ed8c09d53f2d29c01bc8ec37e9..5146ac6792d86d284668ddbb0dba747b67e77bd0 100644 --- a/transcoder_all/layer_42_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_42_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_42_width_16k_l0_small_affine/config.json b/transcoder_all/layer_42_width_16k_l0_small_affine/config.json index d0bbf9eab887ce1d996e7b729450f551101499e1..e6217b82d098e2cfc32213179774381e3da23e63 100644 --- a/transcoder_all/layer_42_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_42_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_42_width_262k_l0_small/config.json b/transcoder_all/layer_42_width_262k_l0_small/config.json index 6966440ee30de6ab45b66d2f3d420a0a750becb8..bcbce4f8c49d7b9173be9c14908d416da2ab0847 100644 --- a/transcoder_all/layer_42_width_262k_l0_small/config.json +++ b/transcoder_all/layer_42_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_42_width_262k_l0_small_affine/config.json b/transcoder_all/layer_42_width_262k_l0_small_affine/config.json index 21423cbbcd100cfcbb9ed4aa58397f14b3a4339c..3d69d17a3f68793744dc08261c1e3528af6ed19a 100644 --- a/transcoder_all/layer_42_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_42_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_big/config.json b/transcoder_all/layer_44_width_16k_l0_big/config.json index 5db3788764673bab24302dd35958023fe2c8c1fe..7f5c289b832134d3cdd2ebaddd484c4c538fde65 100644 --- a/transcoder_all/layer_44_width_16k_l0_big/config.json +++ b/transcoder_all/layer_44_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_45_width_16k_l0_big_affine/config.json b/transcoder_all/layer_45_width_16k_l0_big_affine/config.json index 3bea3bab820349f6a677096618de92db6a92860d..093b773bd76728769951f90e0ef661d1c7bbb15d 100644 --- a/transcoder_all/layer_45_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_45_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_46_width_16k_l0_big/config.json b/transcoder_all/layer_46_width_16k_l0_big/config.json index 1df1c76ad72f4532cdf499b6562bd8d6945d5913..af13e5056850c9ea8269cd81b5aaa0b3ac7b8554 100644 --- a/transcoder_all/layer_46_width_16k_l0_big/config.json +++ b/transcoder_all/layer_46_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_46_width_16k_l0_big_affine/config.json b/transcoder_all/layer_46_width_16k_l0_big_affine/config.json index 88a8db23dd25f0cec399db703c93ce813ea2678a..563319d4b28620a0407e7924a31a972304b447b2 100644 --- a/transcoder_all/layer_46_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_46_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_46_width_262k_l0_small/config.json b/transcoder_all/layer_46_width_262k_l0_small/config.json index 92201d49d756a956249fd217771cf20a1ceff675..2fe7569ad65d2b6f4d9cfacb4dc017352a50a680 100644 --- a/transcoder_all/layer_46_width_262k_l0_small/config.json +++ b/transcoder_all/layer_46_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_47_width_16k_l0_big/config.json b/transcoder_all/layer_47_width_16k_l0_big/config.json index 3dd0836b9ef786eccee9b406b0b98025526e65d3..d80e62b72438e5ee59e828c9ce6762260c3a00fd 100644 --- a/transcoder_all/layer_47_width_16k_l0_big/config.json +++ b/transcoder_all/layer_47_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_47_width_16k_l0_small/config.json b/transcoder_all/layer_47_width_16k_l0_small/config.json index 0cdb396e02615e17c4bd8613630f556b79d7d573..e3e81b214c48a5036f8dd2c3d5fefc9cbd267e26 100644 --- a/transcoder_all/layer_47_width_16k_l0_small/config.json +++ b/transcoder_all/layer_47_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_47_width_262k_l0_big/config.json b/transcoder_all/layer_47_width_262k_l0_big/config.json index 80b7c0c0cd40e9de6d265e30e2138a8a086e7787..75333fded5f553f5fee1618972a0057a4449ae9a 100644 --- a/transcoder_all/layer_47_width_262k_l0_big/config.json +++ b/transcoder_all/layer_47_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_47_width_262k_l0_big_affine/config.json b/transcoder_all/layer_47_width_262k_l0_big_affine/config.json index 0ffd653db93fbac7ee41c76d65c798f6a0a4fc5d..07ff20ed7d1694cb84089283595bd867c63f9cac 100644 --- a/transcoder_all/layer_47_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_47_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_47_width_262k_l0_small/config.json b/transcoder_all/layer_47_width_262k_l0_small/config.json index ae406200fa8bd1725533260d2f2535dd005a5b94..65319ab3b17ddbc1072087a5073f6bbe5fc382a4 100644 --- a/transcoder_all/layer_47_width_262k_l0_small/config.json +++ b/transcoder_all/layer_47_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small/config.json b/transcoder_all/layer_4_width_16k_l0_small/config.json index e7d8bfa66ec1db33ac715685f6df0ae258f99109..60a43268258548c36386aa43c5cb846a210aee11 100644 --- a/transcoder_all/layer_4_width_16k_l0_small/config.json +++ b/transcoder_all/layer_4_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 12, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_big/config.json b/transcoder_all/layer_4_width_262k_l0_big/config.json index 34a02462820e67dd24853df2ca6a097c8336c393..dccd78b4226c43610495d00149709055e287ccdd 100644 --- a/transcoder_all/layer_4_width_262k_l0_big/config.json +++ b/transcoder_all/layer_4_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 75, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_big_affine/config.json b/transcoder_all/layer_4_width_262k_l0_big_affine/config.json index a8cec0a51bcccf6d5c8973ed10590e0907e6ef53..d46514505a8db937e7b31071e85b1a22564405ee 100644 --- a/transcoder_all/layer_4_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_4_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 75, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small/config.json b/transcoder_all/layer_5_width_16k_l0_small/config.json index f61cf3545011233cee3fe88dfb2c9ce086c3ee3b..b0147fef29de4763bbac4e249a60561ef4fda4d5 100644 --- a/transcoder_all/layer_5_width_16k_l0_small/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 13, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_big/config.json b/transcoder_all/layer_5_width_262k_l0_big/config.json index a7753fca6da4744186546982f4865df0fbd8f809..1eec13f6d3e34203b4d0f578132b9367adafbeb8 100644 --- a/transcoder_all/layer_5_width_262k_l0_big/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 78, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big/config.json b/transcoder_all/layer_6_width_16k_l0_big/config.json index 86cc0f9717a094f71d1dd13f173f4bf698245f09..15d619e4eb37a46230f17b17379277b683bc3761 100644 --- a/transcoder_all/layer_6_width_16k_l0_big/config.json +++ b/transcoder_all/layer_6_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 82, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json index 0ffa4066192eaaa488c1da042e44fb4b40570e20..81f53b6e544eb9270058abb2ea2b499f6d77c778 100644 --- a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 86, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big_affine/config.json b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json index daf17db85ae95fa9b102e8b48568911bed7c899a..e24942473247f17679239231d91d07e4fe81fde1 100644 --- a/transcoder_all/layer_7_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 86, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small/config.json b/transcoder_all/layer_7_width_262k_l0_small/config.json index 5005788dfdb474b16dadb2f20deb621aa0a62500..1a7131e051cbcf8387588298a7635e7a1a8477c9 100644 --- a/transcoder_all/layer_7_width_262k_l0_small/config.json +++ b/transcoder_all/layer_7_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 14, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_small_affine/config.json b/transcoder_all/layer_8_width_16k_l0_small_affine/config.json index a3e6125edcfa3bb1a2ad694f7dfc48245d170e96..2198d44cb7539ad5b961e7a71860663af5d6035c 100644 --- a/transcoder_all/layer_8_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_8_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 15, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/config.json b/transcoder_all/layer_8_width_262k_l0_big/config.json index 8f2d7380addb9a7ffcf93e8828671fddaa64ce78..142d30853237c31850a4b797e61f37a5026864c9 100644 --- a/transcoder_all/layer_8_width_262k_l0_big/config.json +++ b/transcoder_all/layer_8_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 90, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big_affine/config.json b/transcoder_all/layer_9_width_262k_l0_big_affine/config.json index e63ee8b6e4bb33dd2215753a4da00f124817d8b8..e6dcca9b2f0d5451279bef1012eddf1a40774115 100644 --- a/transcoder_all/layer_9_width_262k_l0_big_affine/config.json +++ b/transcoder_all/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-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 93, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file