diff --git a/transcoder/layer_17_width_16k_l0_big/config.json b/transcoder/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b40b96fdf6bfb22a7296bca5c5eaea9e29df885 --- /dev/null +++ b/transcoder/layer_17_width_16k_l0_big/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": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_16k_l0_big_affine/config.json b/transcoder/layer_17_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..31bbf7bc4f3122db08755e441759de2a8cbe2479 --- /dev/null +++ b/transcoder/layer_17_width_16k_l0_big_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": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_16k_l0_medium_affine/config.json b/transcoder/layer_17_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8a0a4192e26e9552e1fa35c353c91d3866eb7f37 --- /dev/null +++ b/transcoder/layer_17_width_16k_l0_medium_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": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_16k_l0_small/config.json b/transcoder/layer_17_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c3c034578db7221c5897542f95c142debcd27cf5 --- /dev/null +++ b/transcoder/layer_17_width_16k_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": 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/layer_17_width_262k_l0_big/config.json b/transcoder/layer_17_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ab83c7b4a42a152b9776f35722a56dcb7075b8bc --- /dev/null +++ b/transcoder/layer_17_width_262k_l0_big/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": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_262k_l0_big_affine/config.json b/transcoder/layer_17_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d9a2b4c15e7bfc22963154a14108b9ce50d27ed1 --- /dev/null +++ b/transcoder/layer_17_width_262k_l0_big_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_262k_l0_medium/config.json b/transcoder/layer_17_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..81539f89da8e06b1c78e8c342cef907012d73bac --- /dev/null +++ b/transcoder/layer_17_width_262k_l0_medium/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": 60, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_262k_l0_medium_affine/config.json b/transcoder/layer_17_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..541c76fc3eddd36a26847b41aac677553a1c8d3c --- /dev/null +++ b/transcoder/layer_17_width_262k_l0_medium_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_262k_l0_small_affine/config.json b/transcoder/layer_17_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b9ec0a8a679c392e125923657c29bd270cf95be9 --- /dev/null +++ b/transcoder/layer_17_width_262k_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": 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/layer_17_width_65k_l0_big/config.json b/transcoder/layer_17_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a465bf939066d26f99e7b0e32c9732b808531cf9 --- /dev/null +++ b/transcoder/layer_17_width_65k_l0_big/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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_65k_l0_big_affine/config.json b/transcoder/layer_17_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..51b5c5f833cdcfce232e7583660fffbe02b9636b --- /dev/null +++ b/transcoder/layer_17_width_65k_l0_big_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_65k_l0_medium/config.json b/transcoder/layer_17_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..892de47a89cca023424373e232d92fbef3a6df44 --- /dev/null +++ b/transcoder/layer_17_width_65k_l0_medium/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": 65536, + "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/layer_17_width_65k_l0_medium_affine/config.json b/transcoder/layer_17_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0e3ca4b0d4539e7190bf61cc3ff1b74df0ad979 --- /dev/null +++ b/transcoder/layer_17_width_65k_l0_medium_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_17_width_65k_l0_small/config.json b/transcoder/layer_17_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3c7814a5256a2c457a421bd080f4b0b365fbff11 --- /dev/null +++ b/transcoder/layer_17_width_65k_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": 65536, + "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/layer_17_width_65k_l0_small_affine/config.json b/transcoder/layer_17_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c8135b6de7969f10e805e8246c046cb4d8069bf1 --- /dev/null +++ b/transcoder/layer_17_width_65k_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": 65536, + "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/layer_22_width_16k_l0_big/config.json b/transcoder/layer_22_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e31371cfd40e39a71d0333b4402bbf759a362820 --- /dev/null +++ b/transcoder/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": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_16k_l0_medium/config.json b/transcoder/layer_22_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..862e492a6d7196030ca2181fc27ed31844cf304c --- /dev/null +++ b/transcoder/layer_22_width_16k_l0_medium/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": 60, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_16k_l0_small/config.json b/transcoder/layer_22_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d36d368ac2195220f0f6944605152784ed06a3e --- /dev/null +++ b/transcoder/layer_22_width_16k_l0_small/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": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_16k_l0_small_affine/config.json b/transcoder/layer_22_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d95fa2136bc541e11685f141adba91950e16cdf3 --- /dev/null +++ b/transcoder/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/layer_22_width_262k_l0_big/config.json b/transcoder/layer_22_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0afc233b0c5b505789ae4d6b5bb1c2c52e6acd09 --- /dev/null +++ b/transcoder/layer_22_width_262k_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_262k_l0_big_affine/config.json b/transcoder/layer_22_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4f82eba401446e59970e541825f6ffe47dc223cf --- /dev/null +++ b/transcoder/layer_22_width_262k_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_262k_l0_medium/config.json b/transcoder/layer_22_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d367357a1e89f7f9ffecfcf402c69ada25a91718 --- /dev/null +++ b/transcoder/layer_22_width_262k_l0_medium/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": 262144, + "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/layer_22_width_262k_l0_medium_affine/config.json b/transcoder/layer_22_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7c42c092d5b81cb7f52141e7e7e8a14b29cbce1e --- /dev/null +++ b/transcoder/layer_22_width_262k_l0_medium_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_262k_l0_small_affine/config.json b/transcoder/layer_22_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e1a2fcebcf6a11e513117340b4cea8d419ed0991 --- /dev/null +++ b/transcoder/layer_22_width_262k_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": 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/layer_22_width_65k_l0_big/config.json b/transcoder/layer_22_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d2f4e6938614d75c5b418b896bfedcc7c1a801f3 --- /dev/null +++ b/transcoder/layer_22_width_65k_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_65k_l0_big_affine/config.json b/transcoder/layer_22_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e8031679248f83af4c7bb042d7a2a6fcd249d37 --- /dev/null +++ b/transcoder/layer_22_width_65k_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_65k_l0_medium/config.json b/transcoder/layer_22_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ba625df6609f4766bf7d3bf6914bf2fe7dfdc961 --- /dev/null +++ b/transcoder/layer_22_width_65k_l0_medium/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": 65536, + "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/layer_22_width_65k_l0_medium_affine/config.json b/transcoder/layer_22_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d1920d90b867c97ae727236304a07c9da7510cc1 --- /dev/null +++ b/transcoder/layer_22_width_65k_l0_medium_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_22_width_65k_l0_small/config.json b/transcoder/layer_22_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dd43b61766f464a7dc8b136cb38eacddeb7c49e5 --- /dev/null +++ b/transcoder/layer_22_width_65k_l0_small/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": 65536, + "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/layer_22_width_65k_l0_small_affine/config.json b/transcoder/layer_22_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fd55a952b553d428a9360470e6337dfd8fa90560 --- /dev/null +++ b/transcoder/layer_22_width_65k_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": 65536, + "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/layer_29_width_16k_l0_big/config.json b/transcoder/layer_29_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1baed9021293305d524614f6bc17050269555fae --- /dev/null +++ b/transcoder/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": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_16k_l0_big_affine/config.json b/transcoder/layer_29_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..208c85a430bc79b11d9c509660f5994e8d8005b6 --- /dev/null +++ b/transcoder/layer_29_width_16k_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": 16384, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_16k_l0_medium/config.json b/transcoder/layer_29_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df2e19c73240d7fd1b0b08519344abf41afda43a --- /dev/null +++ b/transcoder/layer_29_width_16k_l0_medium/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": 60, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_16k_l0_medium_affine/config.json b/transcoder/layer_29_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a29fa14ed8218dc965a6b47ba9283b10d6e1cb7a --- /dev/null +++ b/transcoder/layer_29_width_16k_l0_medium_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": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_16k_l0_small/config.json b/transcoder/layer_29_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aeba6f5b15967b53cdf3873fdcba03d778b68633 --- /dev/null +++ b/transcoder/layer_29_width_16k_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": 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/layer_29_width_16k_l0_small_affine/config.json b/transcoder/layer_29_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6faa19c5d064f81c2fb48dd506dcc0bde96c1421 --- /dev/null +++ b/transcoder/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/layer_29_width_262k_l0_big/config.json b/transcoder/layer_29_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dd60a93de6e0ec72e7ab386da1cdab3260efa82a --- /dev/null +++ b/transcoder/layer_29_width_262k_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_262k_l0_big_affine/config.json b/transcoder/layer_29_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b359e3e5c91f02064cd5ffd2e4fb97e6c6b7b84d --- /dev/null +++ b/transcoder/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": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_262k_l0_medium/config.json b/transcoder/layer_29_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..008a07044f0c6c7fff10c46fa1ccc2ee1160b9b9 --- /dev/null +++ b/transcoder/layer_29_width_262k_l0_medium/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": 60, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_262k_l0_medium_affine/config.json b/transcoder/layer_29_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f5fbb8da1f498ea5186410fa4537cd8e0b4a726 --- /dev/null +++ b/transcoder/layer_29_width_262k_l0_medium_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": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_262k_l0_small/config.json b/transcoder/layer_29_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..14d1415c389d42df265e3cc1d8be91bd93015643 --- /dev/null +++ b/transcoder/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/layer_29_width_65k_l0_big/config.json b/transcoder/layer_29_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..302055e4eb7d5ffe6b63f7262ba94b0d17740fc3 --- /dev/null +++ b/transcoder/layer_29_width_65k_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_65k_l0_big_affine/config.json b/transcoder/layer_29_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..60fd7074f331a3291dc8f5ad85b16ec3691e41a1 --- /dev/null +++ b/transcoder/layer_29_width_65k_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_65k_l0_medium/config.json b/transcoder/layer_29_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..764c7df2e2d66b3dd2aaef4088fd6a04fe73f4e4 --- /dev/null +++ b/transcoder/layer_29_width_65k_l0_medium/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": 65536, + "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/layer_29_width_65k_l0_medium_affine/config.json b/transcoder/layer_29_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07bc9c40962da5c3277868f7ab35c72329e7ec1c --- /dev/null +++ b/transcoder/layer_29_width_65k_l0_medium_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_29_width_65k_l0_small/config.json b/transcoder/layer_29_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1ebc8bdf164f73391024b108f3d03265f132e131 --- /dev/null +++ b/transcoder/layer_29_width_65k_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": 65536, + "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/layer_29_width_65k_l0_small_affine/config.json b/transcoder/layer_29_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..817015dc30dcce47083b4fa514a9c92679237dd8 --- /dev/null +++ b/transcoder/layer_29_width_65k_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": 65536, + "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/layer_9_width_16k_l0_big/config.json b/transcoder/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df9649d1abf84212008f02e6ddf737035f2927cc --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_big/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": 131, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_big_affine/config.json b/transcoder/layer_9_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c23c79a692c9dfd9b9dce3d618ea8d17069a5595 --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_big_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": 131, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium/config.json b/transcoder/layer_9_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9d2b5122085b57ec7fb8a0e9997744732d5b229d --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_medium/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": 53, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium_affine/config.json b/transcoder/layer_9_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8474225d2d993685c1dce6b9a21c5f1f86fec97f --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_medium_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": 53, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_small/config.json b/transcoder/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b922f85b253af166cf5a1d8342cfc4a79247f02f --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_small/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": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big/config.json b/transcoder/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b9f1c378218b1aee0444d4399ee3e3e3fcc816c3 --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_big/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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big_affine/config.json b/transcoder/layer_9_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7ab540e19bc70a40011567ce2367f4bec6ca6a3a --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_big_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium/config.json b/transcoder/layer_9_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..23e39ed537ccc7db088494a10a24e62aaa664dc3 --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_medium/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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium_affine/config.json b/transcoder/layer_9_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a2a2fe0f91b46404aa9adbdfe8e43de214c75c08 --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_medium_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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small/config.json b/transcoder/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8bea0cdef324dd185b4b8d5e36644f066d98c0b8 --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_small/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": 262144, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small_affine/config.json b/transcoder/layer_9_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d3b67c47ac4d557ce450cd2848c4dc5c0cee0a1 --- /dev/null +++ b/transcoder/layer_9_width_262k_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": 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/layer_9_width_65k_l0_big/config.json b/transcoder/layer_9_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a1233306db1023ad22d70004409e6981a53f52a7 --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_big/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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_big_affine/config.json b/transcoder/layer_9_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5d0b47df9a9156b67604fbb6c9cb4d8978d63aca --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_big_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 131, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium/config.json b/transcoder/layer_9_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..564bb79e840a7cad4101b7d415491cddea848fa9 --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_medium/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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium_affine/config.json b/transcoder/layer_9_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a119b05e3638d94675b2edfd1c9bfbfb5f8d3d9f --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_medium_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small/config.json b/transcoder/layer_9_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d28b9edadfbec1d98ac5a99880340c41c792a9f3 --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_small/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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small_affine/config.json b/transcoder/layer_9_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..001e3fc5656b3bf48885bed26857560dc423d98c --- /dev/null +++ b/transcoder/layer_9_width_65k_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": 65536, + "model_name": "google/gemma-3-4b-pt", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file