diff --git a/transcoder_all/layer_0_width_16k_l0_small/config.json b/transcoder_all/layer_0_width_16k_l0_small/config.json index 4f18f3f231791cec0480a37c830fee497b5ac81a..22d0620643bd21ccd48d959de9e6e9eaa36d6bd1 100644 --- a/transcoder_all/layer_0_width_16k_l0_small/config.json +++ b/transcoder_all/layer_0_width_16k_l0_small/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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 10, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_big/config.json b/transcoder_all/layer_0_width_262k_l0_big/config.json index 085ec9aae95a4c3158f68c30b9be3284a48036d1..83c86afe88d77c0d023d1d7271419097f8ce3ab7 100644 --- a/transcoder_all/layer_0_width_262k_l0_big/config.json +++ b/transcoder_all/layer_0_width_262k_l0_big/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": 60, - "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/config.json b/transcoder_all/layer_10_width_16k_l0_big/config.json index a15c97e250bdbe243ff1fdb7933fc93cdc18dffc..08ecbaa812e4a8fa4d114d30e9f8ee0027c97fc4 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-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_10_width_262k_l0_big_affine/config.json b/transcoder_all/layer_10_width_262k_l0_big_affine/config.json index c3a4dbd2356aa1f3ff0cd54e577b45d5a04757ed..9584a59933f1d6ed4a6ae4152c9c0e6d2d8a3dbd 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-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_262k_l0_small_affine/config.json b/transcoder_all/layer_10_width_262k_l0_small_affine/config.json index c737500b4a87643c4de752c3d7abb0e856a59799..b7f94c7c795ebdfa5af97971f6ade6c5a0290901 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-it", + "model_name": "google/gemma-3-12b-it", "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_big_affine/config.json b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json index 108513afb8816453e4535bb6cb0d7e4ba7ef7024..d5a310070dd810524474a71851241b1a644bb507 100644 --- a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 101, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_small/config.json b/transcoder_all/layer_11_width_16k_l0_small/config.json index d2241b00145be678f8495ca0aeebda772457c8e0..5336f91fe64e8a4f030e7a18d7284a8c91809a85 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-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_11_width_16k_l0_small_affine/config.json b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json index eb7baa6212c8b85b2b05413b793537a00a862be9..a24c7fabe7839c78ea33208262bae7658455176f 100644 --- a/transcoder_all/layer_11_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 16, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_small/config.json b/transcoder_all/layer_12_width_262k_l0_small/config.json index 2f8469ad09df7c1824a43c4a615fb2db5763f065..f475626c91b299908e3972433dc67e5c206063d1 100644 --- a/transcoder_all/layer_12_width_262k_l0_small/config.json +++ b/transcoder_all/layer_12_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_all/layer_13_width_16k_l0_small/config.json b/transcoder_all/layer_13_width_16k_l0_small/config.json index eae00d93cd631084199e5a1f464984274f1d6e5b..3cead698173568ca38c54b0e4f6367d54308bbfc 100644 --- a/transcoder_all/layer_13_width_16k_l0_small/config.json +++ b/transcoder_all/layer_13_width_16k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_big_affine/config.json b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json index 1701422695b2db4951a5416a820b8de3a1c7a6ab..bd0d900040f7bb7fccda8b93033d85975142e4c1 100644 --- a/transcoder_all/layer_13_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_13_width_262k_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": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "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_15_width_262k_l0_small/config.json b/transcoder_all/layer_15_width_262k_l0_small/config.json index 9a3dabdb33bfb07b101325ce8fe68a0f12a9a6dc..79945b3d55cfb8824c7560e1e183b6012fab0c5d 100644 --- a/transcoder_all/layer_15_width_262k_l0_small/config.json +++ b/transcoder_all/layer_15_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 19, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json index 2f2285c9e3a5dc3ab31342b39f1e0ecdadfdb34d..a21088a108d2644f7e7e1a6fd12b0cb46b5165fc 100644 --- a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_18_width_262k_l0_small/config.json b/transcoder_all/layer_18_width_262k_l0_small/config.json index 9c675be919f9574cb1de0e11668880ff5bcb6745..e3a7d867fdaca3c3428b5ca89b79e846f9426750 100644 --- a/transcoder_all/layer_18_width_262k_l0_small/config.json +++ b/transcoder_all/layer_18_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_1_width_16k_l0_big_affine/config.json b/transcoder_all/layer_1_width_16k_l0_big_affine/config.json index d6416271cdf5ffee2254269f3b09da075b54e5c7..185c191ad0ea6ec8eb60407ca3102df415b94565 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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 63, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_big/config.json b/transcoder_all/layer_20_width_16k_l0_big/config.json index bd0942265f7a4c39b4b865b723d26d4296542488..721954faeddd452224752004d640c6498e501b3d 100644 --- a/transcoder_all/layer_20_width_16k_l0_big/config.json +++ b/transcoder_all/layer_20_width_16k_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": 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_20_width_16k_l0_big_affine/config.json b/transcoder_all/layer_20_width_16k_l0_big_affine/config.json index b8444ff4865498fc85bb38e0c87a15c7d7b22ed1..bb0f786eb1e7d171c16dc1ba54f89667848d14ff 100644 --- a/transcoder_all/layer_20_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_20_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_20_width_16k_l0_small/config.json b/transcoder_all/layer_20_width_16k_l0_small/config.json index a43e4521e1d60b559d4c10ba787dd3ab7e41abc7..0a66627de7b766a8390ccd43cd7bc494e3002793 100644 --- a/transcoder_all/layer_20_width_16k_l0_small/config.json +++ b/transcoder_all/layer_20_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_16k_l0_small_affine/config.json b/transcoder_all/layer_21_width_16k_l0_small_affine/config.json index 6c0978db496cedf8036eca5ddfd160670b210301..2fd4a0d44d836b40da854ffce3ca482bdaf3bcb1 100644 --- a/transcoder_all/layer_21_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_21_width_16k_l0_small_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": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_21_width_262k_l0_small_affine/config.json b/transcoder_all/layer_21_width_262k_l0_small_affine/config.json index 77cdb3a17b83a0512fd1b13b7ae9eef17c4d4d18..1d0628522ff230db9c3e967664b8cd4805cf57ca 100644 --- a/transcoder_all/layer_21_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_21_width_262k_l0_small_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": 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_22_width_16k_l0_big/config.json b/transcoder_all/layer_22_width_16k_l0_big/config.json index 5bb737ec776dad95ddc90986006ee0c9f244a415..eab91a27f28afc20f2bd921f385c6de2afb06e10 100644 --- a/transcoder_all/layer_22_width_16k_l0_big/config.json +++ b/transcoder_all/layer_22_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_22_width_262k_l0_small_affine/config.json b/transcoder_all/layer_22_width_262k_l0_small_affine/config.json index fbf3f053d21d25cae8713a9f8a51c473bed851be..25a1f323e7e82e4540f9c59caa43b502fb11a303 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-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_23_width_16k_l0_small/config.json b/transcoder_all/layer_23_width_16k_l0_small/config.json index f1e1f0616f6e7b991fda0e2d81e3c2ed2beafcc4..73e9516a129beb96aa71e9e882c471ba15e68da9 100644 --- a/transcoder_all/layer_23_width_16k_l0_small/config.json +++ b/transcoder_all/layer_23_width_16k_l0_small/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": 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_23_width_16k_l0_small_affine/config.json b/transcoder_all/layer_23_width_16k_l0_small_affine/config.json index 232097d86f252ec273d6987b5a297c91cb55e1f6..8d87d6f682419d29ae31630fab75090fe0a52d15 100644 --- a/transcoder_all/layer_23_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_23_width_16k_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": 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_24_width_16k_l0_big/config.json b/transcoder_all/layer_24_width_16k_l0_big/config.json index ac291a5c9da5840461cf012b7cedb9971437da51..b0bfd1a0409585318b5ba4991f661186d2dae927 100644 --- a/transcoder_all/layer_24_width_16k_l0_big/config.json +++ b/transcoder_all/layer_24_width_16k_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": 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_24_width_262k_l0_big/config.json b/transcoder_all/layer_24_width_262k_l0_big/config.json index 585c84897a43039fb3dfdedc0040f1cae41270b6..45d14c6cee1bd5541d588c488acd7122e2a73440 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-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_24_width_262k_l0_small/config.json b/transcoder_all/layer_24_width_262k_l0_small/config.json index 620491757dde549c346c5ceed6f0007b3c94e9e7..02924274732e7259d43a7539b805e2e3ef58f06e 100644 --- a/transcoder_all/layer_24_width_262k_l0_small/config.json +++ b/transcoder_all/layer_24_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_25_width_16k_l0_big_affine/config.json b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json index d406def4e3c08f323ac48c59876d4fe217bcea19..8fe4bdaeddb7aac228a96c731baccc44d0940ebe 100644 --- a/transcoder_all/layer_25_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_25_width_16k_l0_small_affine/config.json b/transcoder_all/layer_25_width_16k_l0_small_affine/config.json index 01ee346bade34af80beacc33dd67857c244d6bdb..cd17cb3a2d3c32933902281e45040fdc0ca98ec9 100644 --- a/transcoder_all/layer_25_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_25_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_25_width_262k_l0_big/config.json b/transcoder_all/layer_25_width_262k_l0_big/config.json index cba9866eb1cf4f75c4510a71e10547534a33c160..67a216f2ca298da5a44669b8e72ac109198e62c0 100644 --- a/transcoder_all/layer_25_width_262k_l0_big/config.json +++ b/transcoder_all/layer_25_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", "width": 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_26_width_16k_l0_big/config.json b/transcoder_all/layer_26_width_16k_l0_big/config.json index 22e18d7fca9b8034488df2830b118cae6a61cf7c..f4d21eae039faef6189f6453370bf5f0702128f9 100644 --- a/transcoder_all/layer_26_width_16k_l0_big/config.json +++ b/transcoder_all/layer_26_width_16k_l0_big/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": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_big_affine/config.json b/transcoder_all/layer_26_width_16k_l0_big_affine/config.json index b65080a3f52189130bb1cb3f307e2baf071675ce..d5e8f04311520eee7ffa6a0bb3111b07dfdbe59a 100644 --- a/transcoder_all/layer_26_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_26_width_16k_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": 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_26_width_262k_l0_small_affine/config.json b/transcoder_all/layer_26_width_262k_l0_small_affine/config.json index 79cbb32a68ef5ceb2e9fa1c26aedd4a8d3eceda4..2b0f93da496a2e6c56acaea1c806f0abdd80fd8f 100644 --- a/transcoder_all/layer_26_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_26_width_262k_l0_small_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": 20, - "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_affine/config.json b/transcoder_all/layer_27_width_262k_l0_big_affine/config.json index 24fa4f7bf095c507c20f3ba16f285f5f8cd18e58..7d07045ed551e5d5d74d91f6b18e0091fe9cb9a7 100644 --- a/transcoder_all/layer_27_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_27_width_262k_l0_big_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": 120, - "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/config.json b/transcoder_all/layer_28_width_16k_l0_big/config.json index 0db9a0f4cee7890856e2832202560f7d4dea34c3..fed9b51fc3475a82ecdbcccb04cce46b4e042d1b 100644 --- a/transcoder_all/layer_28_width_16k_l0_big/config.json +++ b/transcoder_all/layer_28_width_16k_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": 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_28_width_16k_l0_small_affine/config.json b/transcoder_all/layer_28_width_16k_l0_small_affine/config.json index b4b3a876d02cdc957996e87f75feff5869fdd29b..2de493cb8e33456d48ade72524edeafe939997c9 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-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_262k_l0_big_affine/config.json b/transcoder_all/layer_28_width_262k_l0_big_affine/config.json index cfb542192ac38e30b3638704fc5ad935755e62e2..917da71aba6400d625db9c0cee71c9c117292f4b 100644 --- a/transcoder_all/layer_28_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_28_width_262k_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": 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_28_width_262k_l0_small/config.json b/transcoder_all/layer_28_width_262k_l0_small/config.json index 8c344df926981d42576c24d2dfb960d0844dcc4a..cb1a9b8f967cbd68270593a56ff1698d0415813f 100644 --- a/transcoder_all/layer_28_width_262k_l0_small/config.json +++ b/transcoder_all/layer_28_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.28.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.28.post_feedforward_layernorm.output", "width": 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_28_width_262k_l0_small_affine/config.json b/transcoder_all/layer_28_width_262k_l0_small_affine/config.json index 4a2ea1c8ca9e1cd0fda73432c188cfe3aba4f181..48bf421568618d50f0e0459df4b30925f76c1efa 100644 --- a/transcoder_all/layer_28_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_28_width_262k_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": 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_29_width_16k_l0_small_affine/config.json b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json index 44b2d05f856aa9791679f8965cdc18446664c46a..5f3095a3ab226f63f1913b312d1a7d96315aaead 100644 --- a/transcoder_all/layer_29_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_29_width_262k_l0_small/config.json b/transcoder_all/layer_29_width_262k_l0_small/config.json index 81f0ab3e3f31ee0eea52e49a375a1248f5f1d8b3..92b0457d88d05d8272aed3d57f3ec74337ea9973 100644 --- a/transcoder_all/layer_29_width_262k_l0_small/config.json +++ b/transcoder_all/layer_29_width_262k_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": 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_2_width_16k_l0_small_affine/config.json b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json index 602d881f5df20c4cdf5e7bc3da6cb33c925af88e..754c2273dbf3d61ca370a4d01481ff12b0fc83b9 100644 --- a/transcoder_all/layer_2_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_2_width_16k_l0_small_affine/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-it", + "model_name": "google/gemma-3-12b-it", "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_2_width_262k_l0_big_affine/config.json b/transcoder_all/layer_2_width_262k_l0_big_affine/config.json index 2452a0ae1c81965fd9a5251ff54d039826e9f332..9098b286f53e6c07f8c4c537ab6ff88896bc89ee 100644 --- a/transcoder_all/layer_2_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_2_width_262k_l0_big_affine/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": 67, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_big/config.json b/transcoder_all/layer_30_width_16k_l0_big/config.json index 7920752006150c6564dff81420bac4492c058831..60ab7b4f1ec3baa826e9e6717464c183eb7c4602 100644 --- a/transcoder_all/layer_30_width_16k_l0_big/config.json +++ b/transcoder_all/layer_30_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_30_width_16k_l0_small_affine/config.json b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json index 9c7ae71ab048c20247b1548b21281fe55989627a..c7d5bf5032e2c416504ac50ee90cdb89e4d9f6f6 100644 --- a/transcoder_all/layer_30_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_30_width_262k_l0_big/config.json b/transcoder_all/layer_30_width_262k_l0_big/config.json index b050f4c6610c7fa28e74d737dfb53cb9ed3a60f2..900e7b99bca6f053fc8b67d144c61b490b2fd8af 100644 --- a/transcoder_all/layer_30_width_262k_l0_big/config.json +++ b/transcoder_all/layer_30_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", "width": 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_31_width_16k_l0_big_affine/config.json b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json index 2772a7559a45e2d5d09803349b3b75d848c3710e..2f8f53d13ffd24b1f93d935cf3366bc5510f2d89 100644 --- a/transcoder_all/layer_31_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_big/config.json b/transcoder_all/layer_31_width_262k_l0_big/config.json index 2ef8b42c14c4a5e66bf443e9b41dd47b757e5e0a..d36412310eea09156b8560be53a6e2921b281811 100644 --- a/transcoder_all/layer_31_width_262k_l0_big/config.json +++ b/transcoder_all/layer_31_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_31_width_262k_l0_small/config.json b/transcoder_all/layer_31_width_262k_l0_small/config.json index 2894467cc0ffb0c55fdff65211794549b1ac8f27..7e4349451348c3c4e06569b432424b2e8b7a4064 100644 --- a/transcoder_all/layer_31_width_262k_l0_small/config.json +++ b/transcoder_all/layer_31_width_262k_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": 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_32_width_262k_l0_big_affine/config.json b/transcoder_all/layer_32_width_262k_l0_big_affine/config.json index 827c252074c72fdc56b3be7143a93ed14bf0f9ee..a7d536c97412ace63664356d653e90c05ed90c48 100644 --- a/transcoder_all/layer_32_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_32_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_32_width_262k_l0_small/config.json b/transcoder_all/layer_32_width_262k_l0_small/config.json index 8a13788f1cf689fccf442924d56c813c80ee796f..6c953545070c9618591c4a14bf38120132e63173 100644 --- a/transcoder_all/layer_32_width_262k_l0_small/config.json +++ b/transcoder_all/layer_32_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", "width": 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_33_width_262k_l0_big_affine/config.json b/transcoder_all/layer_33_width_262k_l0_big_affine/config.json index bac3aee4def68e9f3e1283eb70ff75a675f4b9fe..a0d4314b0ee8285a680282d43b607f93f00aacfb 100644 --- a/transcoder_all/layer_33_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_33_width_262k_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": 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_34_width_16k_l0_small/config.json b/transcoder_all/layer_34_width_16k_l0_small/config.json index 019b94e0290f26895bb503d9ec1f28c63744a243..8e7e73f964f0f05eff71f79b3b5c2b080c4b9c5b 100644 --- a/transcoder_all/layer_34_width_16k_l0_small/config.json +++ b/transcoder_all/layer_34_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_34_width_262k_l0_big/config.json b/transcoder_all/layer_34_width_262k_l0_big/config.json index d9a134cde0539a77114e04fcf5c0e5e3f58623b1..fb4a98f5c0b962ac5e46912130b2d361fcb6723f 100644 --- a/transcoder_all/layer_34_width_262k_l0_big/config.json +++ b/transcoder_all/layer_34_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", "width": 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_35_width_16k_l0_big/config.json b/transcoder_all/layer_35_width_16k_l0_big/config.json index 620fb3515c688f0bc6447f3cd8a181322300e86a..9cf728be9b3093cf367d0839ea228b55f887fd0d 100644 --- a/transcoder_all/layer_35_width_16k_l0_big/config.json +++ b/transcoder_all/layer_35_width_16k_l0_big/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": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_35_width_16k_l0_small/config.json b/transcoder_all/layer_35_width_16k_l0_small/config.json index 92905833f1a983ddf9266177f2a54955a845f9d6..c9b7daa7149adbc01a3268361a7a74347f110df0 100644 --- a/transcoder_all/layer_35_width_16k_l0_small/config.json +++ b/transcoder_all/layer_35_width_16k_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": 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_36_width_262k_l0_big_affine/config.json b/transcoder_all/layer_36_width_262k_l0_big_affine/config.json index 9cbbb650fe4c1dc4a575d01f2eeb4f4a42fafb7e..da66dc98067b0dd4db99c45310d5a54115e84b6c 100644 --- a/transcoder_all/layer_36_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_36_width_262k_l0_big_affine/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": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_small/config.json b/transcoder_all/layer_37_width_16k_l0_small/config.json index a737acd44e415f0ee3191e1b970a9807731be01c..fb084fee6d2af275512534846f0308b0e53fdc38 100644 --- a/transcoder_all/layer_37_width_16k_l0_small/config.json +++ b/transcoder_all/layer_37_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_small_affine/config.json b/transcoder_all/layer_37_width_262k_l0_small_affine/config.json index d30ccad0d521ec17fa8cc23044aea0175d7a88eb..3fcf7fd71a679b282b324374bbd8a2b2b38f8c90 100644 --- a/transcoder_all/layer_37_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_37_width_262k_l0_small_affine/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": 20, - "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/config.json b/transcoder_all/layer_38_width_262k_l0_big/config.json index b1efe4b97b4d16888147b050a33048853622c575..6e0d7867c9247398399443683408e288c69a9b83 100644 --- a/transcoder_all/layer_38_width_262k_l0_big/config.json +++ b/transcoder_all/layer_38_width_262k_l0_big/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-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_16k_l0_big_affine/config.json b/transcoder_all/layer_39_width_16k_l0_big_affine/config.json index 3c311725312c786bec073e7944e1ce8c9d1e32d6..a4326bd2e19c3e77f83d39951c11c360338fe727 100644 --- a/transcoder_all/layer_39_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_39_width_16k_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": 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_262k_l0_small/config.json b/transcoder_all/layer_39_width_262k_l0_small/config.json index 09ee3fb6238b2805f4cfd409d26637fd048ffc85..df342763e49c6e10a631dbd02ec03eaea42cbabb 100644 --- a/transcoder_all/layer_39_width_262k_l0_small/config.json +++ b/transcoder_all/layer_39_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_3_width_262k_l0_big_affine/config.json b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json index a0aaa1ba5b3e62f40289424c587b42db99c126fd..92f2b43f0623fd54ecee2bac7d6574291e4f79cf 100644 --- a/transcoder_all/layer_3_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 71, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small/config.json b/transcoder_all/layer_3_width_262k_l0_small/config.json index 662ecb626f24fd20ca48dec3315c4af7636b6723..75335911eac9b00280f94767581753032090d9bb 100644 --- a/transcoder_all/layer_3_width_262k_l0_small/config.json +++ b/transcoder_all/layer_3_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_40_width_16k_l0_small/config.json b/transcoder_all/layer_40_width_16k_l0_small/config.json index b140da977626436d3dfd9001d6bf6f4a73b2d124..cf400dc4a7e7ea9cb5ae46f79e1a962a7a3afd5f 100644 --- a/transcoder_all/layer_40_width_16k_l0_small/config.json +++ b/transcoder_all/layer_40_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_40_width_16k_l0_small_affine/config.json b/transcoder_all/layer_40_width_16k_l0_small_affine/config.json index 8814c98322a1ee2550334d28d7595c34f35241cc..d6a3145cad11c3e6e0e9f09b70c52f7993ab46f0 100644 --- a/transcoder_all/layer_40_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_40_width_16k_l0_small_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": 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_16k_l0_small/config.json b/transcoder_all/layer_41_width_16k_l0_small/config.json index 661161ba0983d3f2183ef416535e14c39a925d73..cb7cbc513b2d54391392a7c89fba31c0cd48c52a 100644 --- a/transcoder_all/layer_41_width_16k_l0_small/config.json +++ b/transcoder_all/layer_41_width_16k_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": 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_41_width_262k_l0_big/config.json b/transcoder_all/layer_41_width_262k_l0_big/config.json index 3db539815ef88ce05c84bf80c98abb056781ca29..786e97b4d55ea5be0b598422f397e9a7a49c8d11 100644 --- a/transcoder_all/layer_41_width_262k_l0_big/config.json +++ b/transcoder_all/layer_41_width_262k_l0_big/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": 120, - "affine_connection": false + "affine_connection": false, + "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 ab177881afb19a6307516efb626055fc4c104918..efee0eb805c13719f289a3746aa41c2d9dadbed0 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-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_262k_l0_small/config.json b/transcoder_all/layer_41_width_262k_l0_small/config.json index 0d325c8bc5f89383c4541158ed1883e106dbd414..292f34d65cddd9a42c1e1f92bb25742fa3196d80 100644 --- a/transcoder_all/layer_41_width_262k_l0_small/config.json +++ b/transcoder_all/layer_41_width_262k_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": 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_42_width_16k_l0_small_affine/config.json b/transcoder_all/layer_42_width_16k_l0_small_affine/config.json index 5a6608a5196a6a1c7e2be5b421f24a6cff19d6d3..f8088c4a6e4e6dff02ed65f5b9eaf8f3a8d116dd 100644 --- a/transcoder_all/layer_42_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_42_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_262k_l0_big_affine/config.json b/transcoder_all/layer_42_width_262k_l0_big_affine/config.json index dab1b45206c63c542efb277deb2fe829a2008b6d..433a001d3d8a58a6368ef4b782b56b6f1ecb604d 100644 --- a/transcoder_all/layer_42_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_42_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", "width": 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_43_width_16k_l0_small/config.json b/transcoder_all/layer_43_width_16k_l0_small/config.json index c1d50682ba6644399663647c177fe6947b2dfbd5..2c9e1b5f2088442183c99e93d3f4785d59b922b3 100644 --- a/transcoder_all/layer_43_width_16k_l0_small/config.json +++ b/transcoder_all/layer_43_width_16k_l0_small/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": 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_44_width_16k_l0_big/config.json b/transcoder_all/layer_44_width_16k_l0_big/config.json index 51abe50a84eae73443ba0d634eac4764fc7a1aa9..07ff53c6321c0ce55969e4942a7e836ba0351548 100644 --- a/transcoder_all/layer_44_width_16k_l0_big/config.json +++ b/transcoder_all/layer_44_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_44_width_262k_l0_big/config.json b/transcoder_all/layer_44_width_262k_l0_big/config.json index 346c849ccdf4b8ee30bd1c7f07bb76d2a46c991c..89c967be3c4703d64977b391345677c31e898f5f 100644 --- a/transcoder_all/layer_44_width_262k_l0_big/config.json +++ b/transcoder_all/layer_44_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", "width": 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_45_width_16k_l0_small_affine/config.json b/transcoder_all/layer_45_width_16k_l0_small_affine/config.json index 4b1e486388ea17ed55284951f26f79f89dd22f67..13b3d0485bf4d28d1dd951e2d223677818ffa016 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-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_45_width_262k_l0_small/config.json b/transcoder_all/layer_45_width_262k_l0_small/config.json index 1a32d7593257759edf0521e34abfc98ba050ef01..deca57d6b8e90a14b48e6d4f48755e08196911d0 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-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_46_width_16k_l0_small/config.json b/transcoder_all/layer_46_width_16k_l0_small/config.json index 63fc72b358803342175a69ced7776f793922ac0f..13e54408e3a3bcac7bd829f4dc1d448cd49364e2 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-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_46_width_16k_l0_small_affine/config.json b/transcoder_all/layer_46_width_16k_l0_small_affine/config.json index 7f8dffbdf01b9676f771a67ab0074b11390204c1..2f219035ed006b41db02e84dcb8d7eef0feb21ef 100644 --- a/transcoder_all/layer_46_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_46_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_46_width_262k_l0_small/config.json b/transcoder_all/layer_46_width_262k_l0_small/config.json index dae48f55c7bed1c31429a8c07dfcd449f26097e4..28f43c94de99f294f039fca240b004ba406597e1 100644 --- a/transcoder_all/layer_46_width_262k_l0_small/config.json +++ b/transcoder_all/layer_46_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_47_width_16k_l0_big_affine/config.json b/transcoder_all/layer_47_width_16k_l0_big_affine/config.json index e1cf4089d1199d94b99c1d698036a307cff8d8a4..2c2c0386277b383fc0bd5eef931788348964ca32 100644 --- a/transcoder_all/layer_47_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_47_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 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_47_width_16k_l0_small/config.json b/transcoder_all/layer_47_width_16k_l0_small/config.json index 14b75b44fc702d9c2bd3cacd7f4ad2ef8b59d8b5..d742df55f2c97e285159d82a57abdfa64bf4b986 100644 --- a/transcoder_all/layer_47_width_16k_l0_small/config.json +++ b/transcoder_all/layer_47_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-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_47_width_262k_l0_big_affine/config.json b/transcoder_all/layer_47_width_262k_l0_big_affine/config.json index d9f2fb497692af7e558d59ac7b2a41f6ef439155..fe65d8d8992f3113724b6c67ecc77d7593bb1201 100644 --- a/transcoder_all/layer_47_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_47_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_4_width_16k_l0_big_affine/config.json b/transcoder_all/layer_4_width_16k_l0_big_affine/config.json index 23812a3f61c46ea3bcfda6d83926750e2375ff1e..cfe637d84f4659756b5f8c7f2c7ec7b87b6e65de 100644 --- a/transcoder_all/layer_4_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_4_width_16k_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": 16384, - "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_16k_l0_small/config.json b/transcoder_all/layer_4_width_16k_l0_small/config.json index f88b7965e5a3d62f4585869f529c00d95a5f3f43..f6f7c13343defd0c9ae89693d226b0b69606e495 100644 --- a/transcoder_all/layer_4_width_16k_l0_small/config.json +++ b/transcoder_all/layer_4_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 12, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json index df21520290296f0ea42154291bf06024a097b03f..9d2a530d6d48b99fa1edef6c5f0b6ccb30184ea7 100644 --- a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_4_width_16k_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": 16384, - "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_5_width_16k_l0_big/config.json b/transcoder_all/layer_5_width_16k_l0_big/config.json index 83050fc2286707953a3b0c7bacb19784dddc91d2..17b84dd854016cb5d278c94b07c0352c5101ed9e 100644 --- a/transcoder_all/layer_5_width_16k_l0_big/config.json +++ b/transcoder_all/layer_5_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 78, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_small/config.json b/transcoder_all/layer_5_width_262k_l0_small/config.json index 6237edd8ea3f092fe48f10c0545a2148583a3957..dfd852d4b2e41c12218dd90bc5b743bb281cf382 100644 --- a/transcoder_all/layer_5_width_262k_l0_small/config.json +++ b/transcoder_all/layer_5_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 13, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_small_affine/config.json b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json index 1a3e8cd38def7a1470acb3df91350c7a704e9038..36bb1be604c9fd840c19f8af075d4378949ed6ed 100644 --- a/transcoder_all/layer_5_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-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_16k_l0_small/config.json b/transcoder_all/layer_6_width_16k_l0_small/config.json index b4104d464532ea2007bc5a0527c716e057f44c92..a7587c4294358b82df0017601b7d2d06dcd9f4e9 100644 --- a/transcoder_all/layer_6_width_16k_l0_small/config.json +++ b/transcoder_all/layer_6_width_16k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_small/config.json b/transcoder_all/layer_6_width_262k_l0_small/config.json index cc7a8b86ab40bcdbc57e4e021ac81298febf2530..8bdd9b897dc834b35c099d09446d96ee2dba55bd 100644 --- a/transcoder_all/layer_6_width_262k_l0_small/config.json +++ b/transcoder_all/layer_6_width_262k_l0_small/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": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big/config.json b/transcoder_all/layer_7_width_16k_l0_big/config.json index ed630d2d02491c41c526f43cc81a7d1eb633da0c..e4a4e8fe09ae609a0a254f513b9796eb76b916cd 100644 --- a/transcoder_all/layer_7_width_16k_l0_big/config.json +++ b/transcoder_all/layer_7_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 86, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json index 6aac8d6af8ac12c560ccc46208cb22465879a698..3fdbef74c9090b3680b98c4c7a73f1b06c7069b5 100644 --- a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 86, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_small_affine/config.json b/transcoder_all/layer_7_width_16k_l0_small_affine/config.json index dd19c639fdc2967cf4dd5adccaff11a1618eb2f6..ed186fa3a1e39f1c0204bbbaac83cbcf69ab710b 100644 --- a/transcoder_all/layer_7_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_7_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 14, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_big_affine/config.json b/transcoder_all/layer_8_width_16k_l0_big_affine/config.json index 657aa42ae1d4c56937e82b7865098d8aa8a8944a..b2bb97fbb74be990c674baff6fab3ea7150adfb0 100644 --- a/transcoder_all/layer_8_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_8_width_16k_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": 16384, - "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 diff --git a/transcoder_all/layer_9_width_16k_l0_big/config.json b/transcoder_all/layer_9_width_16k_l0_big/config.json index 80011620e55ce0de1dd3249aaac3b8174a6bdd3d..d6ade5ff3b0922556267436ea299b0acfbd99287 100644 --- a/transcoder_all/layer_9_width_16k_l0_big/config.json +++ b/transcoder_all/layer_9_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 93, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_big_affine/config.json b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json index 453931304977a43d286041165d14da3e213af0fc..06377f202e44b7030daa9eaa221353a01657080c 100644 --- a/transcoder_all/layer_9_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 93, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_small_affine/config.json b/transcoder_all/layer_9_width_16k_l0_small_affine/config.json index db0ae7b57ead3518d7823778d56a206c32bb75aa..37d28ddc871bb5bd669a659176ef1829e95aeb35 100644 --- a/transcoder_all/layer_9_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_9_width_16k_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": 16384, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 15, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big_affine/config.json b/transcoder_all/layer_9_width_262k_l0_big_affine/config.json index cafae46b07f619641dcf853b684f807b2d6c5247..cc4f760382e36a9b383d7345eb57448ab637d197 100644 --- a/transcoder_all/layer_9_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_9_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-12b-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 93, - "affine_connection": true + "affine_connection": true, + "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 d5f3d7c91e78cb6caeac001e7f31150fd9f726ba..2d4bf389ba81e9d8ceebbc50a3c25445b85cefaf 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-it", + "model_name": "google/gemma-3-12b-it", "architecture": "jump_relu", "l0": 15, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file