diff --git a/transcoder_all/layer_0_width_16k_l0_big_affine/config.json b/transcoder_all/layer_0_width_16k_l0_big_affine/config.json index fe5ac5bb0841ddc9212d2e625876bea88e660530..96907a9180047e8753321d6ef90c8b35521f2ea8 100644 --- a/transcoder_all/layer_0_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_0_width_16k_l0_big_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": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 60, - "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/config.json b/transcoder_all/layer_10_width_16k_l0_big/config.json index 4fb90ebc6016cab214fafa175250633d8e8351c4..e19fd6491abadb7d542b08d30792002e24b66ab4 100644 --- a/transcoder_all/layer_10_width_16k_l0_big/config.json +++ b/transcoder_all/layer_10_width_16k_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": 16384, - "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_10_width_16k_l0_big_affine/config.json b/transcoder_all/layer_10_width_16k_l0_big_affine/config.json index 605b0db3b86b0ea72d2673ae886ab035ac1b1347..facdf253ecf53735c9afd828cf62c3b960829c61 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-pt", + "model_name": "google/gemma-3-12b-pt", "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_262k_l0_big_affine/config.json b/transcoder_all/layer_10_width_262k_l0_big_affine/config.json index 20204355ba6e03ff8b98bd5274bf0c75827a78a5..a498a0a921b21b6d75d082113794e6aed7929b00 100644 --- a/transcoder_all/layer_10_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_10_width_262k_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": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "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_262k_l0_small_affine/config.json b/transcoder_all/layer_10_width_262k_l0_small_affine/config.json index e8ec26016bee7a2774b837fedbdec64e1e8e8111..7666c70eb5456e7fdfa134b10e31b7863525b6fc 100644 --- a/transcoder_all/layer_10_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_10_width_262k_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": 262144, - "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_11_width_16k_l0_small/config.json b/transcoder_all/layer_11_width_16k_l0_small/config.json index ec2b6b2f049839035f73b997d327949a1cf4d9e8..1ab088aed21054cf98e1471aae0db93c36a12ab1 100644 --- a/transcoder_all/layer_11_width_16k_l0_small/config.json +++ b/transcoder_all/layer_11_width_16k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_big/config.json b/transcoder_all/layer_11_width_262k_l0_big/config.json index 90527c7ddb2ae12e6442f2007f11207342f105b7..ee66996000e2bec72cf62e38d0ca21c64070d63f 100644 --- a/transcoder_all/layer_11_width_262k_l0_big/config.json +++ b/transcoder_all/layer_11_width_262k_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": 262144, - "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_12_width_16k_l0_big_affine/config.json b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json index 6446a80c2896334eaf8d9ddc9231b735b84ffe51..b57391ba0885e7e435668bc366ef87dc674c4c3a 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-pt", + "model_name": "google/gemma-3-12b-pt", "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_16k_l0_small/config.json b/transcoder_all/layer_12_width_16k_l0_small/config.json index 20724a9c7aad8f2872a310d566cd25b4ae4c8a29..518dcc5367d45cc0880b6d03fa5e2f50d4a19af5 100644 --- a/transcoder_all/layer_12_width_16k_l0_small/config.json +++ b/transcoder_all/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-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_big_affine/config.json b/transcoder_all/layer_13_width_16k_l0_big_affine/config.json index f7f3202ef71f3582a9545666d727bb0e4721ca27..e47b0c70d9f0f3684c0b211e928c5fcf24b04c3d 100644 --- a/transcoder_all/layer_13_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_13_width_16k_l0_big_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": 108, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big/config.json b/transcoder_all/layer_14_width_16k_l0_big/config.json index 399da43ca4704d276b614784742bde0defaca1ff..719a19c101847de2d3956d78e9856ce780abffca 100644 --- a/transcoder_all/layer_14_width_16k_l0_big/config.json +++ b/transcoder_all/layer_14_width_16k_l0_big/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_small/config.json b/transcoder_all/layer_14_width_16k_l0_small/config.json index 554220a6148548a633df4510fe8f86f7944645f2..e628ee2204b674c5480054a466228cdd49b5e21d 100644 --- a/transcoder_all/layer_14_width_16k_l0_small/config.json +++ b/transcoder_all/layer_14_width_16k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_small_affine/config.json b/transcoder_all/layer_14_width_262k_l0_small_affine/config.json index cae336e52a0ff629c6f63abaf7c2be9611899b90..f291e7598cd1a40efabb5f066599cca99ba340a8 100644 --- a/transcoder_all/layer_14_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_14_width_262k_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": 262144, - "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_16_width_16k_l0_small_affine/config.json b/transcoder_all/layer_16_width_16k_l0_small_affine/config.json index 86899651e27441adfa05496204ed6a2eb36e42af..55845caa011a4db3281686d56193a5bd4d08765e 100644 --- a/transcoder_all/layer_16_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_16_width_16k_l0_small_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": 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_17_width_16k_l0_small/config.json b/transcoder_all/layer_17_width_16k_l0_small/config.json index 0482b8dc41d90f18d350ab7782e56b76ee007131..dd777f8ac7300715c71ac4ee64e333d4aad656b9 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-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_17_width_262k_l0_big/config.json b/transcoder_all/layer_17_width_262k_l0_big/config.json index fef4cfa2db69f457c02450458d87e868b007dd64..82905016100333b7e81602fcad020da3a0e9904d 100644 --- a/transcoder_all/layer_17_width_262k_l0_big/config.json +++ b/transcoder_all/layer_17_width_262k_l0_big/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": 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_17_width_262k_l0_small/config.json b/transcoder_all/layer_17_width_262k_l0_small/config.json index 768daa3eb6f2fb96314b02c9ffd8aaaf4adf8490..1d412e3f4b9e16e591710c0ce87397e547976560 100644 --- a/transcoder_all/layer_17_width_262k_l0_small/config.json +++ b/transcoder_all/layer_17_width_262k_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": 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_18_width_16k_l0_big/config.json b/transcoder_all/layer_18_width_16k_l0_big/config.json index ece03482936edfc2a591fde17513b73bb74f9f8e..7c419f6487d9d059bfdc4719738582b8ca47c839 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-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_262k_l0_big_affine/config.json b/transcoder_all/layer_19_width_262k_l0_big_affine/config.json index 550340f483cec363a61a8e02e3fcf471f1608bf3..5beed1b504f1d1af9ee871a57497ce244bbdd2d6 100644 --- a/transcoder_all/layer_19_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_19_width_262k_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": 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_1_width_16k_l0_big_affine/config.json b/transcoder_all/layer_1_width_16k_l0_big_affine/config.json index ce23d80f0f81580330073f929a9f3343014c5c5c..f3c544d21eeb9b1bb8dabb2be69264287b713d93 100644 --- a/transcoder_all/layer_1_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_1_width_16k_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": 16384, - "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_1_width_262k_l0_big/config.json b/transcoder_all/layer_1_width_262k_l0_big/config.json index 718aaacd8fc78c7bf1ee6f0055fb119ac9b6162b..9739c1df47baf39c4164396b6f8ba165f3273d3b 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-pt", + "model_name": "google/gemma-3-12b-pt", "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_20_width_262k_l0_big/config.json b/transcoder_all/layer_20_width_262k_l0_big/config.json index 0ab2d0cd539cd484bda6ba0db334ffe44694107d..3f280c8f285a817703263bec5014b3a6f438248b 100644 --- a/transcoder_all/layer_20_width_262k_l0_big/config.json +++ b/transcoder_all/layer_20_width_262k_l0_big/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": 120, - "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 780d45ebd5f00c1332949b6b48628f084680fb7b..9e1441870ed42a6b625be31fcabfea9c399c1545 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-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_262k_l0_small_affine/config.json b/transcoder_all/layer_22_width_262k_l0_small_affine/config.json index 9d72d3500fc3e56f41d26022e51066f18c13a988..ee1826993683614e4f11e5b95d139ad2c66833df 100644 --- a/transcoder_all/layer_22_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_22_width_262k_l0_small_affine/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": 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_24_width_262k_l0_big/config.json b/transcoder_all/layer_24_width_262k_l0_big/config.json index 9a189b629233509678c3014d85c6eb1ece1a9196..ccb6397268bcd35541a91c96e8443b1956ac3a1c 100644 --- a/transcoder_all/layer_24_width_262k_l0_big/config.json +++ b/transcoder_all/layer_24_width_262k_l0_big/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_25_width_262k_l0_small/config.json b/transcoder_all/layer_25_width_262k_l0_small/config.json index dbee4b0cd73b68cb5a86eec4a7a0c6d937dbdd6a..b9df02244aba01bb60f35b1e77568672c88ad4b7 100644 --- a/transcoder_all/layer_25_width_262k_l0_small/config.json +++ b/transcoder_all/layer_25_width_262k_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": 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_27_width_262k_l0_small_affine/config.json b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json index 16d7ac86078641b956edf24eb1e4150876be3d46..a41644fcd4b71e04aee321b23804df8c21ba1733 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-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_28_width_16k_l0_small_affine/config.json b/transcoder_all/layer_28_width_16k_l0_small_affine/config.json index 1e0504a65e6c1cd2057b0789854c4142fb27fda0..055711f6330f827b073ede19d276047e851aae09 100644 --- a/transcoder_all/layer_28_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_28_width_16k_l0_small_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-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_28_width_262k_l0_big/config.json b/transcoder_all/layer_28_width_262k_l0_big/config.json index e88e958c2f65a53fda56fc79d4f66afa758ae1e9..e056b69d752398ba3f7e5a324cedf6f83ae98522 100644 --- a/transcoder_all/layer_28_width_262k_l0_big/config.json +++ b/transcoder_all/layer_28_width_262k_l0_big/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": 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_29_width_16k_l0_small/config.json b/transcoder_all/layer_29_width_16k_l0_small/config.json index b9990154788f6628ff89c6fa5b841d4e91fa0396..f00f69605041e8c6d49f66cf747653268cbbfe46 100644 --- a/transcoder_all/layer_29_width_16k_l0_small/config.json +++ b/transcoder_all/layer_29_width_16k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_29_width_262k_l0_small_affine/config.json b/transcoder_all/layer_29_width_262k_l0_small_affine/config.json index 89f36ed319b603479ef51bc400e9c7cc39e69f90..26481fa6c1ca0b5a89a2b6d7c7a387bb33107057 100644 --- a/transcoder_all/layer_29_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_29_width_262k_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": 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_2_width_16k_l0_big/config.json b/transcoder_all/layer_2_width_16k_l0_big/config.json index abb0058664ab3b7b4bdbe93346954a597a11a8a7..89b1a81fbe3de4b934c219e58bc2097b85ad82f1 100644 --- a/transcoder_all/layer_2_width_16k_l0_big/config.json +++ b/transcoder_all/layer_2_width_16k_l0_big/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": 16384, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 67, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_big/config.json b/transcoder_all/layer_2_width_262k_l0_big/config.json index 75af74e574a0e590953a389c4b1feb4bad500c12..0bae430bb93a3c559b9239dde783d417c36fed05 100644 --- a/transcoder_all/layer_2_width_262k_l0_big/config.json +++ b/transcoder_all/layer_2_width_262k_l0_big/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-pt", + "model_name": "google/gemma-3-12b-pt", "architecture": "jump_relu", "l0": 67, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_small/config.json b/transcoder_all/layer_31_width_16k_l0_small/config.json index 0e9f31a3162cb9ca8c07fcf9c756a926f0dedd26..b7a15a0671b7541f2c930c62941680f2c53f420f 100644 --- a/transcoder_all/layer_31_width_16k_l0_small/config.json +++ b/transcoder_all/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-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_31_width_262k_l0_big_affine/config.json b/transcoder_all/layer_31_width_262k_l0_big_affine/config.json index fcfdc48c0344a63e9f0f18841b6b03cdcdaa1d71..ab2824891c673d61999e5576b33ea16d91e411da 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-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_33_width_16k_l0_big_affine/config.json b/transcoder_all/layer_33_width_16k_l0_big_affine/config.json index 67532d37f871ec9e42095dd3fd91c839ce486137..e830aa164fab6b9af49be4a5b3c31f7ae7acd91c 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-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_38_width_262k_l0_big_affine/config.json b/transcoder_all/layer_38_width_262k_l0_big_affine/config.json index 3a50d6303623bb294047b4e8dbbd9c9a4f80d4f1..cd84366a2890b24a87720bef1720f28fd13ba598 100644 --- a/transcoder_all/layer_38_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_38_width_262k_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": 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_38_width_262k_l0_small/config.json b/transcoder_all/layer_38_width_262k_l0_small/config.json index f2d3efd12ace265dac9cd13f17a9be6c4dc070b3..73c447983181011cefe82a9c65e0ebf91a8b20dc 100644 --- a/transcoder_all/layer_38_width_262k_l0_small/config.json +++ b/transcoder_all/layer_38_width_262k_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": 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_39_width_262k_l0_small_affine/config.json b/transcoder_all/layer_39_width_262k_l0_small_affine/config.json index bf1e80b60dcfbe4cf72bc00c4b790c1039ce4769..870c20478ccc7243278b5d1fdbda06ace7d91a13 100644 --- a/transcoder_all/layer_39_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_39_width_262k_l0_small_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-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_3_width_16k_l0_small_affine/config.json b/transcoder_all/layer_3_width_16k_l0_small_affine/config.json index 944dc887679b0b6a229e013b253450798495fc5f..d5452f715b34ed8d3bc43b7801875064866b05a7 100644 --- a/transcoder_all/layer_3_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_3_width_16k_l0_small_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": 11, - "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 b4e4a935f0400f08cd0b453f1bd78077abde6993..bcc6cb2c9a97c4c02ac37008c3e7a1700deac72f 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-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_41_width_262k_l0_big_affine/config.json b/transcoder_all/layer_41_width_262k_l0_big_affine/config.json index bde5bdf4f8db97fa8b20f86f9d1e6f586accfa4d..88771739ae611dbd793ad9dcc60a7f70078424ba 100644 --- a/transcoder_all/layer_41_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_41_width_262k_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": 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_42_width_16k_l0_big/config.json b/transcoder_all/layer_42_width_16k_l0_big/config.json index d2ad7b38acb0ebb1df5048be89161c3e2bf5aac4..d2bcee73d1e79b03c853e25859c1557e262dbe17 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-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_43_width_262k_l0_big_affine/config.json b/transcoder_all/layer_43_width_262k_l0_big_affine/config.json index 1052cf0040751cd13339e33b44282fa36b58e766..8f2d70638562227029a3e4eba76a83316cdab6be 100644 --- a/transcoder_all/layer_43_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_43_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.43.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_44_width_16k_l0_big_affine/config.json b/transcoder_all/layer_44_width_16k_l0_big_affine/config.json index 037f027e5a5705cec0ebd7ed5d926763adecc374..4f2744c8fd7ea62b0afe7f4a9bb652d889c01490 100644 --- a/transcoder_all/layer_44_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_44_width_16k_l0_big_affine/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": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_small_affine/config.json b/transcoder_all/layer_44_width_16k_l0_small_affine/config.json index f132f950b2638120d07e33a98b9dd15372d344eb..17b76199a5e95ddf677a38eba2a0ae0950ad2c10 100644 --- a/transcoder_all/layer_44_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_44_width_16k_l0_small_affine/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": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_44_width_262k_l0_big_affine/config.json b/transcoder_all/layer_44_width_262k_l0_big_affine/config.json index 71c3319db4a99f99f0a07f37460afe7a8045544e..9358a3c4eee08153671358498aeeb35c9fcd920d 100644 --- a/transcoder_all/layer_44_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_44_width_262k_l0_big_affine/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": 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_44_width_262k_l0_small/config.json b/transcoder_all/layer_44_width_262k_l0_small/config.json index faa2a252f9af553258558dbc644d9ecffbab6f41..ab6362a8d8048d76ce9fce8525418f5052924134 100644 --- a/transcoder_all/layer_44_width_262k_l0_small/config.json +++ b/transcoder_all/layer_44_width_262k_l0_small/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": 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_45_width_16k_l0_small_affine/config.json b/transcoder_all/layer_45_width_16k_l0_small_affine/config.json index 794ffdd07865c55c89e178336ab7336c151cab04..6e6da0da68f8643e0c297083acc0afef9528d3e8 100644 --- a/transcoder_all/layer_45_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_45_width_16k_l0_small_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": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_45_width_262k_l0_small/config.json b/transcoder_all/layer_45_width_262k_l0_small/config.json index c5674702a845febdefd61ec488c80e3c01a3986b..78e4c9bf126066fe908c62253e8c106c52e44ab9 100644 --- a/transcoder_all/layer_45_width_262k_l0_small/config.json +++ b/transcoder_all/layer_45_width_262k_l0_small/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": 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_46_width_16k_l0_small/config.json b/transcoder_all/layer_46_width_16k_l0_small/config.json index fdc941af5401e6e9aea258396fa6478cd2846999..b2eb63d62e8e8001bda83399f4f8b613d6b5c792 100644 --- a/transcoder_all/layer_46_width_16k_l0_small/config.json +++ b/transcoder_all/layer_46_width_16k_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": 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_46_width_262k_l0_big/config.json b/transcoder_all/layer_46_width_262k_l0_big/config.json index e1f0bed13fd0b032748508a76428c0c6f177d8db..7e758e49ea8a48ce017e020fc14c66822b9b6567 100644 --- a/transcoder_all/layer_46_width_262k_l0_big/config.json +++ b/transcoder_all/layer_46_width_262k_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": 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_4_width_16k_l0_big/config.json b/transcoder_all/layer_4_width_16k_l0_big/config.json index a9ee867f2a0583dead464d21b824ef877a15de74..a76ca74688d9eb30bb210890adf80e1bd8b19dac 100644 --- a/transcoder_all/layer_4_width_16k_l0_big/config.json +++ b/transcoder_all/layer_4_width_16k_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": 16384, - "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_8_width_262k_l0_small/config.json b/transcoder_all/layer_8_width_262k_l0_small/config.json index 2d21f4cbe66aca2bb8948123ff88cac6daef61fd..1c8ef417c62fa8a2f9bd0ab06f6ea81b60c096b8 100644 --- a/transcoder_all/layer_8_width_262k_l0_small/config.json +++ b/transcoder_all/layer_8_width_262k_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": 262144, - "model_name": "gemma-v3-12b-pt", + "model_name": "google/gemma-3-12b-pt", "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_9_width_262k_l0_small_affine/config.json b/transcoder_all/layer_9_width_262k_l0_small_affine/config.json index 3bc718cb554dd5e1c4a8690742e3b77861b78226..af852cdc3226a057f26a231b2f60d3efa8345ca5 100644 --- a/transcoder_all/layer_9_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_9_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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