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..a8a6eec2ba316f428d4ea5bd6a12ab41e089872c --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_small/config.json b/transcoder_all/layer_0_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..61d94908a73ee16fbde71c71bf6941304c0565cf --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_small/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big/config.json b/transcoder_all/layer_11_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8886663f5ca94fd00d75939722e6c4687b62ecbe --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 101, + "affine_connection": false +} \ 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..cfadb887c616bb0bf42da46203cae1ff911087c3 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_small_affine/config.json b/transcoder_all/layer_12_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..38fe870a38043c6587d31c5b11f1d80a4edd4298 --- /dev/null +++ b/transcoder_all/layer_12_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_big/config.json b/transcoder_all/layer_13_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..01bb84aa4267eac3b61cbd00bba8c9cc7f00046d --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 108, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_big_affine/config.json b/transcoder_all/layer_13_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df5151c9887421d87b1ab2b41280811c816e3912 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 108, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_small/config.json b/transcoder_all/layer_13_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5335fa467f9800a2ac2f32705bf9c9ff3f574bd9 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ 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..87404b175df24cfe9513aaf1bf9f31ae9b0e351b --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 112, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big/config.json b/transcoder_all/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aa18d56250e1531cdddf09f1f08bb42249d50fbc --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": false +} \ 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..aab6e3dfeb43e2633d07fc1bd045d1f125e4fb25 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": false +} \ 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..2aaf86401bef8172633db8fe512cd9401fd2656d --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_small_affine/config.json b/transcoder_all/layer_15_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..65bb58739aef4661ded2b988dda59dbe941304d2 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 19, + "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..558fae4db3e9c32ff0b998c176ac10ae1471f910 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_16k_l0_small/config.json b/transcoder_all/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7479e91d7b9381be81fca40f74078219a5159ceb --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_16k_l0_small_affine/config.json b/transcoder_all/layer_16_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..97a959d17b80c6d182d894a4abcbb81cd2ff2ab8 --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_small_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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small/config.json b/transcoder_all/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9d12dd6523c38f173516bfc295a830a1e60f3fa5 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_small_affine/config.json b/transcoder_all/layer_17_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..17265f7b81ddc44634bfa0989a30182c710987d1 --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_small_affine/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "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..4d1621705d535ce4ed701cc31c0283d6e14f0435 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_big_affine/config.json b/transcoder_all/layer_17_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f6dd14dede539707ee4e1367752924775429fd7 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_16k_l0_small_affine/config.json b/transcoder_all/layer_18_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..871a4100994a9b9c2617f7e29f33630d9d1a213b --- /dev/null +++ b/transcoder_all/layer_18_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big/config.json b/transcoder_all/layer_18_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..911c7d436b7a67bd230838c27df7ea5e5804999b --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "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..1ae028c080d3c7ade3d1f707e5403d2ab0516ae1 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_big/config.json b/transcoder_all/layer_19_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7eb1b0769bed9b081331ce72d6d1f1c951455608 --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_big/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_small/config.json b/transcoder_all/layer_1_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..18cbd16f1e3d95731a2e7b9f7e4fe9976140b14a --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_small/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ 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..a0992c015220abfd75ceaf88c36ebaad74af0fa3 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "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..43c69c44ad34fe36492878c8372da43f86e41170 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_262k_l0_big_affine/config.json b/transcoder_all/layer_21_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91448211d52fcc1a41efa9c04bf9d28814f5e47b --- /dev/null +++ b/transcoder_all/layer_21_width_262k_l0_big_affine/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_big_affine/config.json b/transcoder_all/layer_22_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4002e9aca85b4a7b7b330ca481b71c9bbb3dc8b --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_big_affine/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_small/config.json b/transcoder_all/layer_22_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9075262cf88d3c88bf6fa0b5627fa53d35bc44a5 --- /dev/null +++ b/transcoder_all/layer_22_width_16k_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": 16384, + "model_name": "gemma-v3-12b-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_big/config.json b/transcoder_all/layer_22_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ffb23e4e51be89fc4c3b694ace86a21843122e37 --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_262k_l0_big_affine/config.json b/transcoder_all/layer_22_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..48bc39cfb44690de1c119a901d13b88a4ba6b13a --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..85ce38ac674bf6ecc15840a91576e9716e060444 --- /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-v3-12b-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/config.json b/transcoder_all/layer_23_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4a56b930019b668110d8d645fdb397ba49c46e6 --- /dev/null +++ b/transcoder_all/layer_23_width_16k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "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..c386728d688a679ce09c07c719a2b0108f23de41 --- /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-v3-12b-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/config.json b/transcoder_all/layer_23_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8687c54379efd1f68c48c7a87a4fd460754f4218 --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_small/config.json b/transcoder_all/layer_23_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..82b55c9b7393d63fd67f78f43fc6981ac11a2f9d --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_big_affine/config.json b/transcoder_all/layer_24_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4449e9ec8ea85f24c45c43060fe9b21e61b47b40 --- /dev/null +++ b/transcoder_all/layer_24_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_small_affine/config.json b/transcoder_all/layer_24_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..78f33414c31479d99de5924d6e1bfc703954d5ca --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_small_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_big/config.json b/transcoder_all/layer_25_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a3132665270a311d452afcc965246d2bee3117ee --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_big/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_262k_l0_small/config.json b/transcoder_all/layer_25_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..17f65d259f89a39c2f9e87ad50b2fe10528b3963 --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..65bb7fea8681d8c71f52f88230ab295e9c2ddda2 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_small/config.json b/transcoder_all/layer_26_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d451d529b0892002f9c1ee084d3efc9f87b7f884 --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_16k_l0_small_affine/config.json b/transcoder_all/layer_27_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7738074da6e8d8a27a892c8cf464e6772e95c1a --- /dev/null +++ b/transcoder_all/layer_27_width_16k_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": 16384, + "model_name": "gemma-v3-12b-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_small/config.json b/transcoder_all/layer_27_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3a7d216207777f16aee6fa8ee5372bdd96f3b2bd --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..7c9ddecbda204d92c2ce608891781a4c4f4e98ff --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_262k_l0_big/config.json b/transcoder_all/layer_28_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e9b119d617864b34507280e8ef2b9f567998ea1b --- /dev/null +++ b/transcoder_all/layer_28_width_262k_l0_big/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_small/config.json b/transcoder_all/layer_29_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c51eab1ba9a566b9eb847d0562eece1035930476 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_262k_l0_small_affine/config.json b/transcoder_all/layer_29_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..de2884fc8d3fe90b3fbddf1cfffe73df4c761aca --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_small_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_big/config.json b/transcoder_all/layer_2_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bb107e9fcb7f670bf3499a4496d6a4cc57e87b8c --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 67, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_small/config.json b/transcoder_all/layer_2_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e08b3ca347cc504998b52ae847035e4d1377e398 --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_big_affine/config.json b/transcoder_all/layer_30_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e6b4e037e96fe289654e06723e34321fc453067 --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_262k_l0_small/config.json b/transcoder_all/layer_30_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2883e0c3ce5877a93434b5dbd16c9272840f0dab --- /dev/null +++ b/transcoder_all/layer_30_width_262k_l0_small/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_small_affine/config.json b/transcoder_all/layer_32_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0d5ca2eaa5b300845ef56067349e13473df6e3e --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_small_affine/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_small_affine/config.json b/transcoder_all/layer_32_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f2013d676f3483b67938ed1504005d8893a0a37 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_16k_l0_small_affine/config.json b/transcoder_all/layer_33_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b5beb64088047bc74a2dab2063b337f0315312da --- /dev/null +++ b/transcoder_all/layer_33_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_big/config.json b/transcoder_all/layer_33_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..17e36281fca8b82f0be237bd07644d9a9f05c315 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "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..afb3810d6198a06347e4fab363c7b0f698e7f75c --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_16k_l0_big_affine/config.json b/transcoder_all/layer_34_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..96207a9e3911dc78d990cd22b8952d1d30564637 --- /dev/null +++ b/transcoder_all/layer_34_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_16k_l0_small_affine/config.json b/transcoder_all/layer_34_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c2167d265519f1a146334d1254176f6007c59a87 --- /dev/null +++ b/transcoder_all/layer_34_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_262k_l0_small/config.json b/transcoder_all/layer_34_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a028f89ae1fc7bed431feeb6373ea76d4a390e2e --- /dev/null +++ b/transcoder_all/layer_34_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_262k_l0_small_affine/config.json b/transcoder_all/layer_34_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c30e9b50cdb8bda330dd275ae8833d162211f9d --- /dev/null +++ b/transcoder_all/layer_34_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "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_affine/config.json b/transcoder_all/layer_35_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..edcf7b9e44019cada448130721f1b3e755b284cc --- /dev/null +++ b/transcoder_all/layer_35_width_16k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_262k_l0_big_affine/config.json b/transcoder_all/layer_35_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6940cec02421d390855e23377462b5f9db7c16a7 --- /dev/null +++ b/transcoder_all/layer_35_width_262k_l0_big_affine/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_262k_l0_small_affine/config.json b/transcoder_all/layer_35_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2df36ed08231c51428b29b90186879837b94dfd2 --- /dev/null +++ b/transcoder_all/layer_35_width_262k_l0_small_affine/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_16k_l0_big_affine/config.json b/transcoder_all/layer_36_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..50ae944ff1097b40171873444e438429cb2ee5e6 --- /dev/null +++ b/transcoder_all/layer_36_width_16k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_big_affine/config.json b/transcoder_all/layer_37_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0fca6d4e10f8da75bb762256b2686ae48c3b7cad --- /dev/null +++ b/transcoder_all/layer_37_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "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..8d33a5788fb74aa24d0178d55de26442c9952f01 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_262k_l0_small/config.json b/transcoder_all/layer_37_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e772051fc6272290a5d7d84f14b97bf1156f3089 --- /dev/null +++ b/transcoder_all/layer_37_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..2880be8b2beb917ff92ac1789a38663b5ad96267 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_16k_l0_small_affine/config.json b/transcoder_all/layer_38_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0675d1a85fecc26623501498f380414efe814dab --- /dev/null +++ b/transcoder_all/layer_38_width_16k_l0_small_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_262k_l0_big_affine/config.json b/transcoder_all/layer_38_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5e051b3e7af306f73716421b0266e1772f0e4350 --- /dev/null +++ b/transcoder_all/layer_38_width_262k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_262k_l0_small_affine/config.json b/transcoder_all/layer_38_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..58de833bbf9080a86584198b88080f5079b0f560 --- /dev/null +++ b/transcoder_all/layer_38_width_262k_l0_small_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_39_width_16k_l0_small/config.json b/transcoder_all/layer_39_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db850b903e8e3a5053fb538e6b7662de6d52f093 --- /dev/null +++ b/transcoder_all/layer_39_width_16k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_39_width_262k_l0_small_affine/config.json b/transcoder_all/layer_39_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..09ae1cac1278a195da9836ac3e62e960bd38d681 --- /dev/null +++ b/transcoder_all/layer_39_width_262k_l0_small_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_small/config.json b/transcoder_all/layer_3_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12833ca7d8cdffb721e7d71f7ba488015d2df0ca --- /dev/null +++ b/transcoder_all/layer_3_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small_affine/config.json b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4950401ff17fea8aa15447d9ebbceb9d0de1e5c2 --- /dev/null +++ b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 11, + "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..72c41b52cdfc7876ec343933c567fac56ecdbf0b --- /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-v3-12b-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_big/config.json b/transcoder_all/layer_41_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9ea462d045b6b00e5d38faafb9945576f40bc039 --- /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-v3-12b-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_big_affine/config.json b/transcoder_all/layer_41_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..61c631beb1dfc8e3ca237778aa685724d9e6f583 --- /dev/null +++ b/transcoder_all/layer_41_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_16k_l0_small/config.json b/transcoder_all/layer_42_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea2328528a8c7056c04e63d7fb408be13fd39393 --- /dev/null +++ b/transcoder_all/layer_42_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_262k_l0_big/config.json b/transcoder_all/layer_42_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..15990258a526de630b8d8d1754a4f86fed27d7ed --- /dev/null +++ b/transcoder_all/layer_42_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_16k_l0_big/config.json b/transcoder_all/layer_43_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a3ff76d18340a097c2f856bb71997e033104a6e6 --- /dev/null +++ b/transcoder_all/layer_43_width_16k_l0_big/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_big/config.json b/transcoder_all/layer_43_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0980e2796bf35cd225562415b7ae53d8252f90d5 --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_big/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_big_affine/config.json b/transcoder_all/layer_44_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..88863924a80bf8cc3f493efcd712da39993b722d --- /dev/null +++ b/transcoder_all/layer_44_width_16k_l0_big_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-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_262k_l0_small/config.json b/transcoder_all/layer_44_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..365954234a341f94df7477cc4c0936b7b51c5bef --- /dev/null +++ b/transcoder_all/layer_44_width_262k_l0_small/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_45_width_262k_l0_big/config.json b/transcoder_all/layer_45_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5141031e18f26b05b840bd8c6455b277b7ca0c46 --- /dev/null +++ b/transcoder_all/layer_45_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_45_width_262k_l0_big_affine/config.json b/transcoder_all/layer_45_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a08a6bdecc758db42cee2f9c2639f5c26d035c0d --- /dev/null +++ b/transcoder_all/layer_45_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_45_width_262k_l0_small_affine/config.json b/transcoder_all/layer_45_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b37c974d64d968691b5a3f41a6adae3635998f61 --- /dev/null +++ b/transcoder_all/layer_45_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_46_width_262k_l0_big/config.json b/transcoder_all/layer_46_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c550e19da1c7b3d004d58767d37d6be51b69d497 --- /dev/null +++ b/transcoder_all/layer_46_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_46_width_262k_l0_big_affine/config.json b/transcoder_all/layer_46_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd06a3129c171cb8e984fe6e3ef7e0a407d7f00f --- /dev/null +++ b/transcoder_all/layer_46_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_47_width_262k_l0_small_affine/config.json b/transcoder_all/layer_47_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4fc024a2962a6bf485cf96d2b8c4843dd8eec66f --- /dev/null +++ b/transcoder_all/layer_47_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_big/config.json b/transcoder_all/layer_4_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc24d845ccb45140f4e0e74bd7c1814233298e1e --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_big/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 75, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_small/config.json b/transcoder_all/layer_4_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9f2405d3f5a909548f0902bf703c30cfc32d80a6 --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small_affine/config.json b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..403b41fc371311954fe3f5ded5edb1d148c299e3 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": true +} \ 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..f032b9b5d3bc28bdd1ac432ac93cd50f217ac482 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 82, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big_affine/config.json b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8627ca0765a3ebaed322070caad5a5a493dabc90 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_big_affine/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 82, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_big/config.json b/transcoder_all/layer_6_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..88d335ddc7ed0186d3591e651496362bb03ce526 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 82, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_big_affine/config.json b/transcoder_all/layer_6_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..722868ab5eb8e9d7071ad8376736c051309f9da7 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_l0_big_affine/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 82, + "affine_connection": true +} \ No newline at end of file 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..67c5e8dc1abdc8a1239e3d11875b40398fa296dd --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": false +} \ No newline at end of file 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..f0b049965e9197a759b4663c1726ea8f3151d461 --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small_affine/config.json b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..294b2e627156e2c0ff5c2de018476aed96c9c809 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small_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-v3-12b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_small_affine/config.json b/transcoder_all/layer_8_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ca62771dd0ca93e72516501121e33cd95c6d8e32 --- /dev/null +++ b/transcoder_all/layer_8_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/config.json b/transcoder_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c5c79815b78de102119dbaa9d6bb234e03019a4c --- /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-v3-12b-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_small/config.json b/transcoder_all/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b578cfb51f538bc677ff298f52e71a3cf46c0f62 --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_small/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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_small/config.json b/transcoder_all/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2b9ca34b87468e0cfe0fd03f2891dac5713de3a4 --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_small/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-v3-12b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file