diff --git a/transcoder/layer_12_width_16k_l0_big/config.json b/transcoder/layer_12_width_16k_l0_big/config.json index dbb90412a708da6c172fbffd20af47f02478dd9d..70effe9fcbcc9b62944331f036f98fd96556d815 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 127, - "affine_connection": false + "affine_connection": false, + "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 958a247fa5723de6458408b74d1df99151cb20c1..374dc16ba1137133f373df877067b2727b66e37c 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 17, - "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 e3174a1ac54417d8c40aff19f165ed24e6483517..7b97208c2bb87b0c944df529e65f66aa4e743feb 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 127, - "affine_connection": true + "affine_connection": true, + "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 db14713aea2267bf4c60daac1db9e28a7e4517e1..9c4f139e7a103b635f18768158cf620328f22100 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 17, - "affine_connection": true + "affine_connection": true, + "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 2d0a0fbddfd76ba5bb99ea86f672615e050a7fcd..ba6b8ad91f708c15c868a4ebadf1ebdda128279a 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 127, - "affine_connection": true + "affine_connection": true, + "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 00edff2bcb772dddc0066027fa183be0191435d5..dd4c2a0658c99b144a8f741b7324e9dcd57dbfad 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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 52, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_16k_l0_medium/config.json b/transcoder/layer_24_width_16k_l0_medium/config.json index 63539b526b360c263a8a39edbf74c098aaa810f5..03f7461ae86147c2ce219907c49ca0ec76e3e020 100644 --- a/transcoder/layer_24_width_16k_l0_medium/config.json +++ b/transcoder/layer_24_width_16k_l0_medium/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_16k_l0_small/config.json b/transcoder/layer_24_width_16k_l0_small/config.json index b698d376cb2c9a98db3b25e906d13fe1902c8bc3..ea4c708426c1cf7090233db53c68132c6818a2e0 100644 --- a/transcoder/layer_24_width_16k_l0_small/config.json +++ b/transcoder/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_medium/config.json b/transcoder/layer_24_width_262k_l0_medium/config.json index f439d63bc72f773e116d81ff7227a296357ee031..74854a39afa4cdfeb40314c99fe3b7ce23809205 100644 --- a/transcoder/layer_24_width_262k_l0_medium/config.json +++ b/transcoder/layer_24_width_262k_l0_medium/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_medium_affine/config.json b/transcoder/layer_24_width_262k_l0_medium_affine/config.json index edb3ccb6cf1978be2b3350e32b17b71622c40447..4be0701a35e532742079e323ef648f1e9c28b7ea 100644 --- a/transcoder/layer_24_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_24_width_262k_l0_medium_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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_small_affine/config.json b/transcoder/layer_24_width_262k_l0_small_affine/config.json index bf27a3f1ec79473f301465e037e6b67103ee33c5..294276e2f09b773e5e2d34acf08ea9249bc7d61a 100644 --- a/transcoder/layer_24_width_262k_l0_small_affine/config.json +++ b/transcoder/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_65k_l0_big_affine/config.json b/transcoder/layer_24_width_65k_l0_big_affine/config.json index dcbbabaa32886de6fff989efd69f4c750719b63f..3600465faccfc9abec39e775aee5620120487712 100644 --- a/transcoder/layer_24_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_24_width_65k_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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_24_width_65k_l0_small/config.json b/transcoder/layer_24_width_65k_l0_small/config.json index a28b0d52f9bea4d30c34b09034af57c12fdbcbdc..67a65089727300b17b3c902d7a8e8b14ccc1c5e1 100644 --- a/transcoder/layer_24_width_65k_l0_small/config.json +++ b/transcoder/layer_24_width_65k_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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_31_width_16k_l0_big/config.json b/transcoder/layer_31_width_16k_l0_big/config.json index a45b8211e6c4b741cef686650bd1e67abc357b78..a5bbd2b22d0a1f4477f60662aee6cae889f47b0c 100644 --- a/transcoder/layer_31_width_16k_l0_big/config.json +++ b/transcoder/layer_31_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 150, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_31_width_16k_l0_small/config.json b/transcoder/layer_31_width_16k_l0_small/config.json index f16c144650fb440f97889bd000bb284807a6c71f..e40778bbae809e2633a997257b46d8e543ad6f9d 100644 --- a/transcoder/layer_31_width_16k_l0_small/config.json +++ b/transcoder/layer_31_width_16k_l0_small/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_31_width_65k_l0_medium/config.json b/transcoder/layer_31_width_65k_l0_medium/config.json index 8a8b51d04c03080d9863d54e34b45cf7403b17b8..86a130181f6884fa67aa5571e52a7517775485a8 100644 --- a/transcoder/layer_31_width_65k_l0_medium/config.json +++ b/transcoder/layer_31_width_65k_l0_medium/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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_31_width_65k_l0_medium_affine/config.json b/transcoder/layer_31_width_65k_l0_medium_affine/config.json index 2c7c9fb21004802fa21da6a4d9b24fd1771d0cda..8d9c2a6a028d016a39a15e7bd597a98c0c91f9b7 100644 --- a/transcoder/layer_31_width_65k_l0_medium_affine/config.json +++ b/transcoder/layer_31_width_65k_l0_medium_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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_41_width_16k_l0_big_affine/config.json b/transcoder/layer_41_width_16k_l0_big_affine/config.json index 7da75e8cedd703fcc4c8cf61e4b43c64d857a973..87cbb2e29a249d4df1ca65c071b0485b4893e01e 100644 --- a/transcoder/layer_41_width_16k_l0_big_affine/config.json +++ b/transcoder/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_41_width_16k_l0_medium/config.json b/transcoder/layer_41_width_16k_l0_medium/config.json index f946072222e3047f68a5e4591bbe04301ffed55c..bc0abdd372e3eab187e3f6a101c531bf07f17a76 100644 --- a/transcoder/layer_41_width_16k_l0_medium/config.json +++ b/transcoder/layer_41_width_16k_l0_medium/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_41_width_262k_l0_medium_affine/config.json b/transcoder/layer_41_width_262k_l0_medium_affine/config.json index 1262a5158cbff1c24d62150b87ae2f454b2c0371..f299bc448404d6edf8d7cfc1ef2d1dc81c33626c 100644 --- a/transcoder/layer_41_width_262k_l0_medium_affine/config.json +++ b/transcoder/layer_41_width_262k_l0_medium_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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 60, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_41_width_65k_l0_big_affine/config.json b/transcoder/layer_41_width_65k_l0_big_affine/config.json index 6c06b0e44320e0128b99b7fd54f293d0b73fe6a6..56c43896b61c7731428cf0889ea33fc6b31c3d05 100644 --- a/transcoder/layer_41_width_65k_l0_big_affine/config.json +++ b/transcoder/layer_41_width_65k_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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 150, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder/layer_41_width_65k_l0_small/config.json b/transcoder/layer_41_width_65k_l0_small/config.json index 97e692f30ad9b1aa48447a7c8268e3204a562293..2dcf555eb14529658998bff824ab4d0eed8a71c2 100644 --- a/transcoder/layer_41_width_65k_l0_small/config.json +++ b/transcoder/layer_41_width_65k_l0_small/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": 65536, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_0_width_262k_l0_small_affine/config.json b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json index 96f9ebc5d13982f624e9c901a226cb0fb6ac549f..5ee55438cf6c043c91f42e07c7f69ee01c4c61b9 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-it", + "model_name": "google/gemma-3-12b-it", "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_big_affine/config.json b/transcoder_all/layer_10_width_16k_l0_big_affine/config.json index 439505d0de875e94f1d465fe8dab7676bc9a6012..698a9c7e4314db323dcec5ae9dd727a13a18c205 100644 --- a/transcoder_all/layer_10_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_10_width_16k_l0_big_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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 97, - "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 c90fdf0556402dba250faa844c2d9bb81d810126..cab062cd13c74d7bd61946dad524d8357c2b5ef0 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-it", + "model_name": "google/gemma-3-12b-it", "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_262k_l0_big/config.json b/transcoder_all/layer_10_width_262k_l0_big/config.json index 98e038f852f2182c6d3841db39d7572a08267c79..b915c899774faf3d53a69ac7e699804f105f7c0c 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-it", + "model_name": "google/gemma-3-12b-it", "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_12_width_16k_l0_big_affine/config.json b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json index 4af22e8b34d48a4ed58737b04bdae28b8dc20ae2..e54678814a5be315a8d8f59a02f10b8ca9a795d1 100644 --- a/transcoder_all/layer_12_width_16k_l0_big_affine/config.json +++ b/transcoder_all/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-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 105, - "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 aebc2ad5b2928b0a305a702412dc04b6481babe5..1f74fd2826cf3a3233c37aa43430f9c8b8eac46b 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-it", + "model_name": "google/gemma-3-12b-it", "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_13_width_16k_l0_small_affine/config.json b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json index e918ccd85db7e763f8c6d21e08b2050c4062b827..17276cc6f276f335f31a084744d4282c30374a9c 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-it", + "model_name": "google/gemma-3-12b-it", "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 03147664bedc420a8ea2b37f5e5738d4cdd6c682..63f52cfba8176e2dc203b74f40566f963600696f 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-it", + "model_name": "google/gemma-3-12b-it", "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_small_affine/config.json b/transcoder_all/layer_14_width_16k_l0_small_affine/config.json index 0e1f541fa59881aebc37cf53f13b8ecf737f5525..17dd7006245b087c127dce2e67e2d4b7ee431709 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-it", + "model_name": "google/gemma-3-12b-it", "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_17_width_16k_l0_big_affine/config.json b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json index cf261fd26cda04554568ab78b99beda410f50e2c..7b35ddfd1dd87657d23faadebe4faf96388f35f9 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-it", + "model_name": "google/gemma-3-12b-it", "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_small/config.json b/transcoder_all/layer_17_width_16k_l0_small/config.json index f65356c3f8d5b558cb3cd3ed53b0cb40be45ea2d..3f1f701b0f72139a2c31c1c55d04b95ff0287069 100644 --- a/transcoder_all/layer_17_width_16k_l0_small/config.json +++ b/transcoder_all/layer_17_width_16k_l0_small/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-it", + "model_name": "google/gemma-3-12b-it", "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_18_width_16k_l0_big/config.json b/transcoder_all/layer_18_width_16k_l0_big/config.json index e90902405aa6e14c70c858c78ade65339a4b6820..51c0b9868a3413933a2b42d91443b04dc7dfedbe 100644 --- a/transcoder_all/layer_18_width_16k_l0_big/config.json +++ b/transcoder_all/layer_18_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_16k_l0_big_affine/config.json b/transcoder_all/layer_18_width_16k_l0_big_affine/config.json index f981aba84ba5960b1a8ece3ae6245cd9f1326a81..9788648edafe009f2ddc1325c96d23cc540711bc 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-it", + "model_name": "google/gemma-3-12b-it", "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_1_width_262k_l0_big/config.json b/transcoder_all/layer_1_width_262k_l0_big/config.json index ec99680478c2abf3123a4d2a7fd3755fe697dbe3..beea74da7180be03fe58dbb0d404e82c3722e8e0 100644 --- a/transcoder_all/layer_1_width_262k_l0_big/config.json +++ b/transcoder_all/layer_1_width_262k_l0_big/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 63, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_big/config.json b/transcoder_all/layer_21_width_16k_l0_big/config.json index 79ea404ee2bbc0f3bfa2b0c90b03d7ffedef69e1..b5ae2d5431bd9a5aef1a89d6401320104cb14e54 100644 --- a/transcoder_all/layer_21_width_16k_l0_big/config.json +++ b/transcoder_all/layer_21_width_16k_l0_big/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-it", + "model_name": "google/gemma-3-12b-it", "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_21_width_16k_l0_big_affine/config.json b/transcoder_all/layer_21_width_16k_l0_big_affine/config.json index 1e2bef34ea97d4ababb511d215a3561c7f3045f3..9a4b0f1d519715be0d6ffe8c2249c04cb32390a6 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-it", + "model_name": "google/gemma-3-12b-it", "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 477ca3b50dee53321766b9afb066c144b4a6106f..b241320f9c7db76efdcad34feb039440101160b9 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-it", + "model_name": "google/gemma-3-12b-it", "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_21_width_262k_l0_small/config.json b/transcoder_all/layer_21_width_262k_l0_small/config.json index 91a08deda01b06dec6baa4a6d990d5a2a3890e90..b6ccc1cf2f2aa6ec9504be6f9cdead1f0fb54318 100644 --- a/transcoder_all/layer_21_width_262k_l0_small/config.json +++ b/transcoder_all/layer_21_width_262k_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_23_width_262k_l0_small_affine/config.json b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json index 8071ab5956d9fec0df072596ecd6951f32b9ff4c..8d6c30503f3bfd26933040b1d0cc755895c95c34 100644 --- a/transcoder_all/layer_23_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_24_width_262k_l0_big_affine/config.json b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json index 1982792aaee5083c092c7981c52a97dd79f2c6d0..f8241b9ae68b6c0af38f4be39ad046a0360e83aa 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-it", + "model_name": "google/gemma-3-12b-it", "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_26_width_16k_l0_small/config.json b/transcoder_all/layer_26_width_16k_l0_small/config.json index 889eb9da69cb5f8658c77d292c884169a99060b5..3799e0e06c2a46a0b187e63da687e4173fc52bb8 100644 --- a/transcoder_all/layer_26_width_16k_l0_small/config.json +++ b/transcoder_all/layer_26_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_26_width_262k_l0_big_affine/config.json b/transcoder_all/layer_26_width_262k_l0_big_affine/config.json index 16eaee1f75fcdd9e30bc1d22a9b5c0f579005fcc..f5cc074fbdb9345bfdcdd26499b94a82828c7987 100644 --- a/transcoder_all/layer_26_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_26_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_27_width_262k_l0_big/config.json b/transcoder_all/layer_27_width_262k_l0_big/config.json index 91d13e225987d0162278619b3214098b46209f23..6acb85bc0ac01b720afa1eb9d64e61f408d33710 100644 --- a/transcoder_all/layer_27_width_262k_l0_big/config.json +++ b/transcoder_all/layer_27_width_262k_l0_big/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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_27_width_262k_l0_small_affine/config.json b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json index c4b3c26fe321897cda3bc847fc617c38f5a64dcc..4e21b5947bbc1780f7efdaa88f3784b099bdaf0b 100644 --- a/transcoder_all/layer_27_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_27_width_262k_l0_small_affine/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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_28_width_16k_l0_big_affine/config.json b/transcoder_all/layer_28_width_16k_l0_big_affine/config.json index d04c8928d5ef8e27c338884a401656ec1cc37832..5f3cc756db855babc59e56c8747a1faa5065be67 100644 --- a/transcoder_all/layer_28_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_28_width_16k_l0_big_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_2_width_262k_l0_small/config.json b/transcoder_all/layer_2_width_262k_l0_small/config.json index ea48da14a8bd255a0469c9c7af160d0d0b0e6df0..3af589074de4bc28f4216b3b585fdac0cf9dc787 100644 --- a/transcoder_all/layer_2_width_262k_l0_small/config.json +++ b/transcoder_all/layer_2_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_30_width_262k_l0_big_affine/config.json b/transcoder_all/layer_30_width_262k_l0_big_affine/config.json index eccce2093c23be92cc21b409b1b089e6e923f605..22eb8a98dc7bfbbf7c09fc3c59e4b6841fc631be 100644 --- a/transcoder_all/layer_30_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_30_width_262k_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_16k_l0_small_affine/config.json b/transcoder_all/layer_31_width_16k_l0_small_affine/config.json index da3d924899b115001940334b8927fa16d1d2106a..85e6f8866b4ac228052b052d66b57b52d27e5a47 100644 --- a/transcoder_all/layer_31_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_31_width_16k_l0_small_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-it", + "model_name": "google/gemma-3-12b-it", "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_262k_l0_big_affine/config.json b/transcoder_all/layer_31_width_262k_l0_big_affine/config.json index fe8a3e943f3662e01a44ff86fc1c3f804d1d236c..29dd75af6f8c9b280fc1d6b7711acbb26565748a 100644 --- a/transcoder_all/layer_31_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_31_width_262k_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_small_affine/config.json b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json index 65b51798eb6908c9ef6314dd67227a6213a9ec71..d8433ab3959d8e77c0b5296110df9dca8369fd74 100644 --- a/transcoder_all/layer_31_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_31_width_262k_l0_small_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_affine/config.json b/transcoder_all/layer_33_width_16k_l0_big_affine/config.json index 80d9ae26ec14980a8d38a247f2cce3d6368d794c..2f1d9d18f0b06b057481b33eebd00db71fc83d4a 100644 --- a/transcoder_all/layer_33_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_33_width_16k_l0_big_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_33_width_16k_l0_small/config.json b/transcoder_all/layer_33_width_16k_l0_small/config.json index 5e43f45b1fabee22c659bc1a88d07c5ac6dafe6f..5487ec5843f7d96727dca5fd638745ff5a0a5888 100644 --- a/transcoder_all/layer_33_width_16k_l0_small/config.json +++ b/transcoder_all/layer_33_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_33_width_262k_l0_small_affine/config.json b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json index 64683ac388c9ae9032472b3192f2881fccc39b07..015f1150948be4f6d4d404a45a6ee6f2a8654aac 100644 --- a/transcoder_all/layer_33_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_33_width_262k_l0_small_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_35_width_16k_l0_small_affine/config.json b/transcoder_all/layer_35_width_16k_l0_small_affine/config.json index e435f1b29c301dff0d5e04f292ba34ec74916b5d..abab13fd26b222bfbad08a91fe1f97fd308bdafb 100644 --- a/transcoder_all/layer_35_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_35_width_16k_l0_small_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-it", + "model_name": "google/gemma-3-12b-it", "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_35_width_262k_l0_small/config.json b/transcoder_all/layer_35_width_262k_l0_small/config.json index fd1025d5d5915f00bafbe726ebd708c42bcc658a..8d73ad49aad9b2fd8e8e3b6b80fa65503f7b9d6c 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-it", + "model_name": "google/gemma-3-12b-it", "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_16k_l0_big/config.json b/transcoder_all/layer_36_width_16k_l0_big/config.json index 1dba1e6e7851526396054edfccf5394a9cce21a8..43ccca6078e6f9c35c5978f9b0275ce2a5e01c59 100644 --- a/transcoder_all/layer_36_width_16k_l0_big/config.json +++ b/transcoder_all/layer_36_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_36_width_262k_l0_small/config.json b/transcoder_all/layer_36_width_262k_l0_small/config.json index 110c98f6f723cefb1f09689b95ea5f363bb302b2..df03b6d69c027235e3be98dcaf2b0238581a494a 100644 --- a/transcoder_all/layer_36_width_262k_l0_small/config.json +++ b/transcoder_all/layer_36_width_262k_l0_small/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-it", + "model_name": "google/gemma-3-12b-it", "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_big/config.json b/transcoder_all/layer_37_width_262k_l0_big/config.json index 42f41303762250492d74a9e3acf9693d29070c5e..f90d1b757b4feff18e459bcb6120f11d52916307 100644 --- a/transcoder_all/layer_37_width_262k_l0_big/config.json +++ b/transcoder_all/layer_37_width_262k_l0_big/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-it", + "model_name": "google/gemma-3-12b-it", "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_38_width_16k_l0_big_affine/config.json b/transcoder_all/layer_38_width_16k_l0_big_affine/config.json index 7cb097f2a1e87e9b4aef9133a459e701b9d9cbe4..338a778740690196d375ab8d7bcf0ac5697b6fb3 100644 --- a/transcoder_all/layer_38_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_38_width_16k_l0_big_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_39_width_16k_l0_big/config.json b/transcoder_all/layer_39_width_16k_l0_big/config.json index 2f609143a02a88a67cdde1805d7b69bc44dc82a7..093083a3227a59234f7f49a4bf5336530cb47be5 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-it", + "model_name": "google/gemma-3-12b-it", "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_affine/config.json b/transcoder_all/layer_39_width_262k_l0_big_affine/config.json index 3694c9d92c0052a837638de349b96bb13335a5e0..b64c47b494792ae56cb13e7c5125475359f88a36 100644 --- a/transcoder_all/layer_39_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_39_width_262k_l0_big_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_40_width_262k_l0_big_affine/config.json b/transcoder_all/layer_40_width_262k_l0_big_affine/config.json index d18450de6f24947a3ea834ecf200490d7a120c93..5fd6fcd590e3cdd85faa33da019ad974d4d63679 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-it", + "model_name": "google/gemma-3-12b-it", "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_small_affine/config.json b/transcoder_all/layer_41_width_16k_l0_small_affine/config.json index cea0ca18d83a9e887e3113c3847a2f9eae3c78c4..1f097e8a852cf87dc0e8ff4496660443a4ce42e3 100644 --- a/transcoder_all/layer_41_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_41_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_41_width_262k_l0_small_affine/config.json b/transcoder_all/layer_41_width_262k_l0_small_affine/config.json index 150d3196190ff135b503c20b88919cdad227af16..46771794fc77faca61b667af1bd2c324541e34cd 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-it", + "model_name": "google/gemma-3-12b-it", "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/config.json b/transcoder_all/layer_42_width_16k_l0_big/config.json index 5837de4f7b6e9cae5e784cf8a708526855ecbc43..b36809cc770f33195c7ead9a08e41f0dfe7341f6 100644 --- a/transcoder_all/layer_42_width_16k_l0_big/config.json +++ b/transcoder_all/layer_42_width_16k_l0_big/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-it", + "model_name": "google/gemma-3-12b-it", "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/config.json b/transcoder_all/layer_45_width_16k_l0_big/config.json index c9f9121978a12580bd9dacd133606392b13cd4b8..6bd48a32071466d37909641dc01947a64ca63159 100644 --- a/transcoder_all/layer_45_width_16k_l0_big/config.json +++ b/transcoder_all/layer_45_width_16k_l0_big/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-it", + "model_name": "google/gemma-3-12b-it", "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_big/config.json b/transcoder_all/layer_47_width_16k_l0_big/config.json index d928cb75dbb7edbd2c67a23d5d5854fefd2c57a4..92e1bb19a06704636c50bccaff34b984f452f2e7 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-it", + "model_name": "google/gemma-3-12b-it", "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_4_width_262k_l0_big_affine/config.json b/transcoder_all/layer_4_width_262k_l0_big_affine/config.json index ea47de3e7fff7816618d5ff399141b5b30a908bb..dc189e8228831f84f416d185d567ab1f580d490a 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-it", + "model_name": "google/gemma-3-12b-it", "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_4_width_262k_l0_small_affine/config.json b/transcoder_all/layer_4_width_262k_l0_small_affine/config.json index 8094251a1860954599c8edf7456203bd95be1e0d..4935ec4795bca549c16a6cd5ab58db0f2e3a4651 100644 --- a/transcoder_all/layer_4_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_4_width_262k_l0_small_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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 12, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_small_affine/config.json b/transcoder_all/layer_6_width_16k_l0_small_affine/config.json index c174f2fa2fbbb521bbdda633e4dbb08648e8901d..620073cff8218bb650a793532d9253c0861560e1 100644 --- a/transcoder_all/layer_6_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_6_width_16k_l0_small_affine/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 13, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_small_affine/config.json b/transcoder_all/layer_6_width_262k_l0_small_affine/config.json index 646c054ab20ce8c4586bd585ca9b4a6b6f37d445..fdb0826e5dc769073e19fbc14070d20d134a892c 100644 --- a/transcoder_all/layer_6_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_6_width_262k_l0_small_affine/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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 13, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_small/config.json b/transcoder_all/layer_8_width_16k_l0_small/config.json index 3f6746a1b18e1a332f2c2d96862c42dd4d85cbf2..0ea69712df825bb5c7699840881e483d892486f2 100644 --- a/transcoder_all/layer_8_width_16k_l0_small/config.json +++ b/transcoder_all/layer_8_width_16k_l0_small/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 15, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big_affine/config.json b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json index 1c68f1bc639e9196e7d385609f49a38c55867124..98a0217159d65c83ae372022d13d0940e118bd72 100644 --- a/transcoder_all/layer_8_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_8_width_262k_l0_big_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 90, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file