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..8a7258f6ffc7cdfff4635989c4d8a511d648dc3c --- /dev/null +++ b/transcoder_all/layer_0_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ 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..8b17626b8044ecbed0d55703806a1632b3f85b2b --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 89, + "affine_connection": false +} \ 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..e3694b67570c1580d74adbb11c1374d440768713 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_small/config.json b/transcoder_all/layer_10_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..988d5b85b186215a23a4f858fb0384abef3158fa --- /dev/null +++ b/transcoder_all/layer_10_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_small/config.json b/transcoder_all/layer_11_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8bf9ab0af2866d5948c53be36cf3483f1ea10c4c --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_small_affine/config.json b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1bdcbc28d56df4bfbc782fb41b0d333917ccfb0c --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_small/config.json b/transcoder_all/layer_11_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..60767c0220be630cfc40e1ce0decd58d95e6e4d0 --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_big_affine/config.json b/transcoder_all/layer_12_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0d6d058e0d76d2bb97aa27ce411451446ea176cf --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 94, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_small/config.json b/transcoder_all/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cc0a31dcd58e0d513d0db7b26a9e20ee30a37d46 --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_small_affine/config.json b/transcoder_all/layer_12_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5f7563a06524f3314bc869265b4b9856b5567f9b --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true +} \ 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..f9af946baa99f3e2a7f04d57a9a200140fbbb622 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ 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..91e00c743cd081c76ce47929b5ebe95020497e40 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big_affine/config.json b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6e29bd629af6ff1b83692b8c76e446b39d195ee0 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 100, + "affine_connection": true +} \ 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..1d973c4068ec03725558e10422feb3f09a6db8e5 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 103, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_small/config.json b/transcoder_all/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8df8d92273715c0baea7759b931e00112809b572 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_small_affine/config.json b/transcoder_all/layer_15_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd67834674c4d0105d7ed01a0a0a376bb98a1e92 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_big_affine/config.json b/transcoder_all/layer_15_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..69c8d119758744d9867216656cef46ad1d67bf24 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 103, + "affine_connection": true +} \ 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..c160b37454550567e852eb82aa0eb3ba94ed3bbd --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 106, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9c87078e5e1c241cd1c75e98e268bcf96667c1f1 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 106, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_big/config.json b/transcoder_all/layer_17_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f42a8f8238accae1fe89e01416176afb92e4f27 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 109, + "affine_connection": false +} \ 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..139f9197c7055db7c064775b134d38547d6beb59 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4dbb58dae5fcdacea63ada40d569846c21ce0a8f --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 112, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_small/config.json b/transcoder_all/layer_18_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4e8481ea6dcbb37c63984fc3f81e8c56fe9f511a --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_small_affine/config.json b/transcoder_all/layer_18_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6901a42b4045bdf154377ed39301d0f0754e6c0e --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_262k_l0_small/config.json b/transcoder_all/layer_19_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3b66a7175c760914e72be5ae23a4c32d91c6e2aa --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_small_affine/config.json b/transcoder_all/layer_1_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1502e4863d1884f947f9d61b1ba6b28c9053206 --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_big_affine/config.json b/transcoder_all/layer_20_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ecd51895504df3afa7ee9c3aeaf7d5fd898996e4 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 118, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_small/config.json b/transcoder_all/layer_20_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..75fb52123ab4d13bbcb346b0c5c99367f293c3f1 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_small/config.json b/transcoder_all/layer_21_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7a3bde7b96b348b9feb1774cdca67a4269b1ec2c --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_262k_l0_small/config.json b/transcoder_all/layer_22_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..88bfe0984683fedb25dca6d9c60c9d0f75fa3060 --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_16k_l0_big_affine/config.json b/transcoder_all/layer_23_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c6f6c9b664c78e0caadee249c243dadc421203e3 --- /dev/null +++ b/transcoder_all/layer_23_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_big_affine/config.json b/transcoder_all/layer_23_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..172be675b952f0396cc74866c5850bdec885d7b6 --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_big_affine/config.json b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dc6f5433a967b6fca1ce85c389f752ae6980d829 --- /dev/null +++ b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..5b32f51a81932c202ef9b64b3c556a206077d251 --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_big/config.json b/transcoder_all/layer_27_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4e4d28f051594a8a25ebfcd8ed56ea9f18a84e0d --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_big_affine/config.json b/transcoder_all/layer_27_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1f4894e88feb32f05c561d847758012a89b89aa --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_small_affine/config.json b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aeaf16c85832a3884d8b14d898fc78895752fe93 --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ 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..fda9ca35e10ad912ac972ac036fe2a1029947926 --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.28.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.28.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..dd50ce55f31f6b3054e75aaad8a062a554198a75 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_262k_l0_big/config.json b/transcoder_all/layer_29_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..52c1984074d0744ade5c59ead5810ab4152d3d0a --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..9a9869d6145259ef6caa17da14f8fd11faa8b6bd --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_big_affine/config.json b/transcoder_all/layer_2_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..51367afc64759f4d4ed84da87b9b406d24000b5e --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 65, + "affine_connection": true +} \ 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..45570fb89df06581c37f79c796ff1fe83beaf8ff --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..e01ff3af3883797d820566e7c0691432c99bfd96 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..1009d7b69ae3f8a17a0093e9513b500f33ef18cc --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..5a4d27ea3c062ed0d1bbef646804ac1d585b4163 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_16k_l0_big/config.json b/transcoder_all/layer_35_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d517e0392f803692688cb2927171146601b9a8d9 --- /dev/null +++ b/transcoder_all/layer_35_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.35.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.35.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_16k_l0_small/config.json b/transcoder_all/layer_36_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..41ff8a83471071612dc1c70cf2a3263233e4c044 --- /dev/null +++ b/transcoder_all/layer_36_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_small/config.json b/transcoder_all/layer_37_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ae7f70d8f369134246935e9d269ff34a3f8d5236 --- /dev/null +++ b/transcoder_all/layer_37_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_small_affine/config.json b/transcoder_all/layer_37_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..76d5070ad731aa5007a88fd7ed020f46b5937761 --- /dev/null +++ b/transcoder_all/layer_37_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_262k_l0_big_affine/config.json b/transcoder_all/layer_37_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3c180d759cec0680e024994a6374a63c0e32f91d --- /dev/null +++ b/transcoder_all/layer_37_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_16k_l0_big/config.json b/transcoder_all/layer_38_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..327fe7456b9b162ea0b9eb7edd2725f4ff7bfa2e --- /dev/null +++ b/transcoder_all/layer_38_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_262k_l0_big/config.json b/transcoder_all/layer_38_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..72e465b6edbfe4dc84090a3a36f036cf83f3a6ce --- /dev/null +++ b/transcoder_all/layer_38_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_39_width_16k_l0_big_affine/config.json b/transcoder_all/layer_39_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..51401013e2589ced3ac55f4536fa691605b6336c --- /dev/null +++ b/transcoder_all/layer_39_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_16k_l0_big/config.json b/transcoder_all/layer_40_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9e783c885ac2f78ecc0eb0b66239156c7558ae82 --- /dev/null +++ b/transcoder_all/layer_40_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_262k_l0_small/config.json b/transcoder_all/layer_40_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e86fd6556fc5f87e8e3ee450255a130b31cb7f27 --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_262k_l0_small_affine/config.json b/transcoder_all/layer_40_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e78e43c6b9fbd853e7ee691b5537b4304f3dea13 --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_41_width_16k_l0_big/config.json b/transcoder_all/layer_41_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..591e2cb6d8e168d4af1680ed36bceb7b867722b3 --- /dev/null +++ b/transcoder_all/layer_41_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_41_width_16k_l0_small/config.json b/transcoder_all/layer_41_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..371ab83ad3c7431618c046942649736b3fbe332b --- /dev/null +++ b/transcoder_all/layer_41_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_41_width_262k_l0_big_affine/config.json b/transcoder_all/layer_41_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e922760ea730896c91b8c32d6b45707ec9a26d71 --- /dev/null +++ b/transcoder_all/layer_41_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_small_affine/config.json b/transcoder_all/layer_43_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db0b75701ac56d85b8086bad88ab945e56bc7603 --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_small_affine/config.json b/transcoder_all/layer_44_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..737cb7d021f250e954a242ef238505d30d9885e1 --- /dev/null +++ b/transcoder_all/layer_44_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_49_width_16k_l0_big_affine/config.json b/transcoder_all/layer_49_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..687a5da441ab5d7236f4e489ffab46119ca2a1c8 --- /dev/null +++ b/transcoder_all/layer_49_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.49.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.49.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e680fa0686895bc9e652eb2ad30fb29ff3510b23 --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_50_width_16k_l0_small_affine/config.json b/transcoder_all/layer_50_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..41b9a41a7681ffcc3f8f9d1cf700f2e6fcef7564 --- /dev/null +++ b/transcoder_all/layer_50_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.50.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.50.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_50_width_262k_l0_small_affine/config.json b/transcoder_all/layer_50_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bbe7460260ae1b9f8134067706915fa6eda0496e --- /dev/null +++ b/transcoder_all/layer_50_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.50.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.50.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_51_width_262k_l0_big/config.json b/transcoder_all/layer_51_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b3de7c25c7177c9625c6e99bdcb808eb8b2cacf3 --- /dev/null +++ b/transcoder_all/layer_51_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.51.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.51.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_52_width_16k_l0_big_affine/config.json b/transcoder_all/layer_52_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fef8bd5ecf4c9a85c40303bdc2507b58faae1548 --- /dev/null +++ b/transcoder_all/layer_52_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.52.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.52.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_52_width_16k_l0_small/config.json b/transcoder_all/layer_52_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cfb936f7594220081d828357f4eb8b3f3ce26b73 --- /dev/null +++ b/transcoder_all/layer_52_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.52.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.52.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_52_width_16k_l0_small_affine/config.json b/transcoder_all/layer_52_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eddb096105fc00c8b2fef3f7e73405dfffd412a --- /dev/null +++ b/transcoder_all/layer_52_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.52.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.52.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_52_width_262k_l0_small/config.json b/transcoder_all/layer_52_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d910a9ac999b1662a112b88ec6d8b0c997c6e6ec --- /dev/null +++ b/transcoder_all/layer_52_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.52.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.52.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_54_width_262k_l0_small_affine/config.json b/transcoder_all/layer_54_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6058b7ad253a97b49a032ae1bfef520be541258d --- /dev/null +++ b/transcoder_all/layer_54_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.54.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.54.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_55_width_16k_l0_big_affine/config.json b/transcoder_all/layer_55_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7e50ef8a46f62278a651e460aa9ea2478518a9c5 --- /dev/null +++ b/transcoder_all/layer_55_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.55.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.55.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_56_width_16k_l0_small_affine/config.json b/transcoder_all/layer_56_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..43bf3d32416b839e0ae6bcb904ca7b2576677f01 --- /dev/null +++ b/transcoder_all/layer_56_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.56.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.56.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_57_width_16k_l0_big_affine/config.json b/transcoder_all/layer_57_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..80284511f23a1b279deed63eab1f46059458b3b5 --- /dev/null +++ b/transcoder_all/layer_57_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.57.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.57.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_57_width_262k_l0_big/config.json b/transcoder_all/layer_57_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c5318bf5bcffd239be175390c26a6e7f8e011f95 --- /dev/null +++ b/transcoder_all/layer_57_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.57.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.57.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_57_width_262k_l0_big_affine/config.json b/transcoder_all/layer_57_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..347384184aa55f52509618113d873a901e6b6f33 --- /dev/null +++ b/transcoder_all/layer_57_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.57.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.57.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_58_width_16k_l0_big/config.json b/transcoder_all/layer_58_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91266833ed42b39665723ae51e321e393bf1c8bf --- /dev/null +++ b/transcoder_all/layer_58_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.58.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.58.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big/config.json b/transcoder_all/layer_6_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a1e6a6e4738c45adfbfd501f220e9e8b9acfb19d --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 77, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big/params.safetensors b/transcoder_all/layer_6_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/transcoder_all/layer_7_width_16k_l0_small/config.json b/transcoder_all/layer_7_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..542ed41f9e7afeb28f091debd4e2384aec3e9a7f --- /dev/null +++ b/transcoder_all/layer_7_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_small/params.safetensors b/transcoder_all/layer_7_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/transcoder_all/layer_7_width_262k_l0_big_affine/config.json b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ebb7eff3b32f413020ccb6782536bc2d5feceb8 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_7_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/transcoder_all/layer_7_width_262k_l0_small/config.json b/transcoder_all/layer_7_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..69b1a99ad77802d5fc81fe9bcada1d2f893811c1 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small/params.safetensors b/transcoder_all/layer_7_width_262k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 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..02bc8edd7735b9e5ec2e943dc8ef6d03576939fa --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 83, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/params.safetensors b/transcoder_all/layer_8_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 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..d4403a48bd1a2950f191f3501353a511d18d8125 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_8_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/transcoder_all/layer_9_width_262k_l0_big_affine/config.json b/transcoder_all/layer_9_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6aa34bf3f85bcf0a5ddde663ce1059bb786351d3 --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_9_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391