diff --git a/resid_post/layer_17_width_16k_l0_big/config.json b/resid_post/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8c8be8a625b0cfe68a6c0ee2c93b1bba0307bfd1 --- /dev/null +++ b/resid_post/layer_17_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_1m_l0_small/config.json b/resid_post/layer_17_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1a16562e791d2f0b2479b15233448f4891bca5e5 --- /dev/null +++ b/resid_post/layer_17_width_1m_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_262k_l0_medium/config.json b/resid_post/layer_17_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cda42700cf141acea43641f7228f9fdf35299e57 --- /dev/null +++ b/resid_post/layer_17_width_262k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_17_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cda42700cf141acea43641f7228f9fdf35299e57 --- /dev/null +++ b/resid_post/layer_17_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_65k_l0_big/config.json b/resid_post/layer_17_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..721539d7475099b58d32d607c679a2a0709c5dbd --- /dev/null +++ b/resid_post/layer_17_width_65k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_65k_l0_medium/config.json b/resid_post/layer_17_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..032e6b3e23bb98b76921e626594c90593a71ab30 --- /dev/null +++ b/resid_post/layer_17_width_65k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_17_width_65k_l0_small/config.json b/resid_post/layer_17_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..073f3ceed19e9807a0dd4a96ec0a445fbb9b6091 --- /dev/null +++ b/resid_post/layer_17_width_65k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_16k_l0_small/config.json b/resid_post/layer_22_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..eaaa5cf14a6accf79719a7e014deefa455dbfebc --- /dev/null +++ b/resid_post/layer_22_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_1m_l0_big/config.json b/resid_post/layer_22_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3bec447824aae6405c1c1aad205a98908ebd5caa --- /dev/null +++ b/resid_post/layer_22_width_1m_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_1m_l0_medium/config.json b/resid_post/layer_22_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7d13a480ab13d24ced9962532940379ce1bf067 --- /dev/null +++ b/resid_post/layer_22_width_1m_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_262k_l0_medium/config.json b/resid_post/layer_22_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6f779c1550a293c29e725b3f6f7948eef4fbd91e --- /dev/null +++ b/resid_post/layer_22_width_262k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_262k_l0_small/config.json b/resid_post/layer_22_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07be5ccb24041acf696e41eb2a2dae6b6317ae3d --- /dev/null +++ b/resid_post/layer_22_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_65k_l0_big/config.json b/resid_post/layer_22_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8ba2a17e39d2a3820ca51f80f0bc426ac48ceb3a --- /dev/null +++ b/resid_post/layer_22_width_65k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_65k_l0_medium/config.json b/resid_post/layer_22_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0d85e4c80442d2863289bf727c05411c9866b7e6 --- /dev/null +++ b/resid_post/layer_22_width_65k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_22_width_65k_l0_small/config.json b/resid_post/layer_22_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3d7a4631fac58cde8a33e02c9f860f3113ad5032 --- /dev/null +++ b/resid_post/layer_22_width_65k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.output", + "hf_hook_point_out": "model.layers.22.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_16k_l0_medium/config.json b/resid_post/layer_29_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cc2f9e03ab7e2be930e78acf168d3e0159459971 --- /dev/null +++ b/resid_post/layer_29_width_16k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_16k_l0_small/config.json b/resid_post/layer_29_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..855ac4b35ad98597bd3bd9ce98a134c5ae48e95b --- /dev/null +++ b/resid_post/layer_29_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_1m_l0_big/config.json b/resid_post/layer_29_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e994805fdcea972e37deb5c6bf38f5d989c45a8d --- /dev/null +++ b/resid_post/layer_29_width_1m_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_1m_l0_medium/config.json b/resid_post/layer_29_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..192c71520565661df50c09b7cb35cafcefb554b3 --- /dev/null +++ b/resid_post/layer_29_width_1m_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_1m_l0_small/config.json b/resid_post/layer_29_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b2df28635ff2f9e167244bd132d688816d961940 --- /dev/null +++ b/resid_post/layer_29_width_1m_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_262k_l0_big/config.json b/resid_post/layer_29_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6357e0ffb958e95383d8f3b0705e5e99f80c50e2 --- /dev/null +++ b/resid_post/layer_29_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_262k_l0_medium/config.json b/resid_post/layer_29_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a133bfe0540e0606ae530c5e71ce15f7e928213 --- /dev/null +++ b/resid_post/layer_29_width_262k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_29_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a133bfe0540e0606ae530c5e71ce15f7e928213 --- /dev/null +++ b/resid_post/layer_29_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_65k_l0_big/config.json b/resid_post/layer_29_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8d619d9a2bf29a853e011cf9ab6f2dea31243fc5 --- /dev/null +++ b/resid_post/layer_29_width_65k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_29_width_65k_l0_medium/config.json b/resid_post/layer_29_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..73ec44b88f2d8324a0d30642db6db80bbfd5f38b --- /dev/null +++ b/resid_post/layer_29_width_65k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.output", + "hf_hook_point_out": "model.layers.29.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_16k_l0_big/config.json b/resid_post/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c6da334c1b2964b7deff52b236a1dad7e0459a8 --- /dev/null +++ b/resid_post/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_16k_l0_medium/config.json b/resid_post/layer_9_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..114ce694c691f4ef27dd8e6698e14771c14c7e69 --- /dev/null +++ b/resid_post/layer_9_width_16k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_1m_l0_medium/config.json b/resid_post/layer_9_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..95270d0b3946ce026739f6901ec7b6b9e47c1120 --- /dev/null +++ b/resid_post/layer_9_width_1m_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 1048576, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_262k_l0_medium/config.json b/resid_post/layer_9_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..268aff25ea6ffd0cc2a01052b0558d7c58d2900a --- /dev/null +++ b/resid_post/layer_9_width_262k_l0_medium/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_65k_l0_big/config.json b/resid_post/layer_9_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a131cd6c462628e9dc32e39f1043d0642baa4707 --- /dev/null +++ b/resid_post/layer_9_width_65k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/resid_post/layer_9_width_65k_l0_small/config.json b/resid_post/layer_9_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84d98a97459ce99ebad016bb3bdf76783f94a84f --- /dev/null +++ b/resid_post/layer_9_width_65k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false, + "type": "sae" +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_16k_l0_big/config.json b/transcoder_all/layer_0_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e8efb5442a92fb34e44f47f3e3a0002fc4b5dd01 --- /dev/null +++ b/transcoder_all/layer_0_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_small/config.json b/transcoder_all/layer_0_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4da56e12874be4197d2961a574089fae7bd754b5 --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 10, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..ebcfa30b8e96a9feead2cae369c87e644a69e2df --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 112, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small/config.json b/transcoder_all/layer_10_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ab0c77c9263552a7c87a0cf3d5f58d42ba86b27f --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e03c9fd6224e2ec23425f04c3ed3ef290d07ab17 --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 118, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..892fba1f9ddf016974189852f1342856ddeae7fc --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 118, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..43b7ad267d2419c727e929ffac579af1097e907e --- /dev/null +++ b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..4feca6111b16a61926f0987961f020c99436b4b4 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..5b128f045fc4e566afc7650ef9a6248c97e1549b --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..768aaaf376298091176ee7295c9b9b45249c66e0 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..821a44f5f0870f351b837718de1ccc7d0d014413 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_big_affine/config.json b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f5d69640f4e1ee0143ede562dbbc7f12030c0907 --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big_affine/config.json b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dce708b92e06b4f2913191faf375ac2981c3672d --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_16k_l0_big/config.json b/transcoder_all/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..64b9ed37c76b6f6e18951a54a23be6db8cefa0c7 --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_16k_l0_big_affine/config.json b/transcoder_all/layer_16_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..31bf97c55388f3cc553b94ab89d26716021eaa0c --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small/config.json b/transcoder_all/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..88bed2036b922bedbf7a505444094c267b2d0145 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_small_affine/config.json b/transcoder_all/layer_17_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..61a168d8f0a29f8c3a0976c10b0444e4e8906494 --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..c141e02b6df6b330159603f5399daeb540def564 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_16k_l0_small/config.json b/transcoder_all/layer_18_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4db5373cc972784e11696c0415103d75880e8aa --- /dev/null +++ b/transcoder_all/layer_18_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_big_affine/config.json b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b266e2c729d8e36615f0c66006d63daf2b84c65b --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..1f43dcf2c246762727fa95efcef5fb436bc50892 --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_small/config.json b/transcoder_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..08996e00a65b8ce24b40bcc8d73d678ed8579aaf --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..454173a2ea92ea54f14054015b7ded9809b0a11f --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..355e590a107dceb814ceb8269106a26749724d63 --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_small_affine/config.json b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..854d7529cb942f1982432c531c1bf4e64fbb0920 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..3af85ddf1bacc1f825798b1354eb1cecc53a7320 --- /dev/null +++ b/transcoder_all/layer_21_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..2d2da41d135b1c2cbf43dd542f18e636b532cac7 --- /dev/null +++ b/transcoder_all/layer_21_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..b9d56e2573ce29d3f80168044ffa8b2fd877d4f0 --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_big_affine/config.json b/transcoder_all/layer_22_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e9319014e06c5c7f337b30c7aac0094bed4bef76 --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_small_affine/config.json b/transcoder_all/layer_22_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d95fa2136bc541e11685f141adba91950e16cdf3 --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..3c0ff183d30d69759ac26163899717604ab2d967 --- /dev/null +++ b/transcoder_all/layer_23_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_big_affine/config.json b/transcoder_all/layer_24_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a996738501641b996268139782cf10a2b3d2f7de --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_small/config.json b/transcoder_all/layer_24_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4145bb59657d5b2229a57ff0b6a7ebb99e41bcbf --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..b77a44a71e9fe6170cbb2606376fe23d15a1de19 --- /dev/null +++ b/transcoder_all/layer_24_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..65ca98148a97df2d3447998a9687097a63baa3e8 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..92fc5683185b299d19c7951672b9c7145c18c020 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_262k_l0_small_affine/config.json b/transcoder_all/layer_25_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..94cc53c33e3e32b05d168d09bbc445ff0b37c48c --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_small/config.json b/transcoder_all/layer_26_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e87ee913aa50bfaab9893bc40b5f3ae7c50e73bc --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_16k_l0_big_affine/config.json b/transcoder_all/layer_28_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8455e5ff71db43886624bfd690eac381c7e35382 --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_16k_l0_small/config.json b/transcoder_all/layer_28_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4a980589315db6af4ae5029d955420ed8dd111fa --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..f5643e0585fd5bea9fb90ae18e8b270318fc12c9 --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..bd5858310e1a0b0983eaf41420ce8d1546764854 --- /dev/null +++ b/transcoder_all/layer_28_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..23d6d7f7bfc2eb8f07bd08d87acca00ccfb9acad --- /dev/null +++ b/transcoder_all/layer_28_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_big/config.json b/transcoder_all/layer_29_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..709a8dfd408de5fd46e7428bbed0aed0a0bb37cb --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_small_affine/config.json b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6faa19c5d064f81c2fb48dd506dcc0bde96c1421 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_262k_l0_big_affine/config.json b/transcoder_all/layer_29_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1eda169bd5ee71ee7d39bc4864a11ec530a33572 --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..14d1415c389d42df265e3cc1d8be91bd93015643 --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..c3500bf2b149e769366ce5a5dd2eadff808ed4ac --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..5838a8595b434fac819d7f5f28eb4084bcb83183 --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..e21b37d50631db3d6c88bcb029a641cf292ed473 --- /dev/null +++ b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..cdf90acc3595e3f8705454fd61e8c7ebfac88eb3 --- /dev/null +++ b/transcoder_all/layer_31_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..726e2930cb30ff8c366fa43080d5cd667a8703a2 --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_262k_l0_small_affine/config.json b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a715442b40209979dd0e5560b8c1cd40fb484141 --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_small/config.json b/transcoder_all/layer_32_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e159fe09ebfa76a2f6cff92e20806e692e26d8e --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..d9e780ac7869b44758c78cb39e7b054d8de77b70 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_small/config.json b/transcoder_all/layer_33_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ae55853ad62509bdd76f583051a1e0a76d321468 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_small_affine/config.json b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ddeb399e89fa911fb0d5906fa0feb13f479c559e --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 20, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..a3f0ca233995e395a3ec47fe74afb75e0551e3b8 --- /dev/null +++ b/transcoder_all/layer_3_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": false, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..fe4b07a3a12e49b80cd2cb9da7765139b5e06bca --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 81, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..015f92ecd292d57439d3ae6be52d59645b3c3222 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small_affine/config.json b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..faf667af6eb3a3dd84b8c08bf27d5c29a9f34d11 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big_affine/config.json b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3286c5d116414ee72b88fdaaebf93144837c39b6 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 91, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_small_affine/config.json b/transcoder_all/layer_6_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e54ef2de16ef737e2e8108d8e9ab8e0f836ad39 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_small_affine/config.json b/transcoder_all/layer_6_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ce5c06887e1014ac006e1378c441801cdf00d727 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big/config.json b/transcoder_all/layer_7_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c8b86a99db1caa3fbf9b30f2f1a25f7ec5d8a2a8 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small_affine/config.json b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c9f1c00b0f03fc826e87d17361d189a189f4ea65 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/config.json b/transcoder_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8f38c5dc74b428405e9f16b74710f8c58b2c7fc6 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 102, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small_affine/config.json b/transcoder_all/layer_8_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..54ce7b37594482c174e550fc6b88559188e5be61 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "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 new file mode 100644 index 0000000000000000000000000000000000000000..a4e437368d462481e1ef2d6817ffcd036a8e832b --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file