diff --git a/transcoder_all/layer_0_width_16k_l0_small_affine/config.json b/transcoder_all/layer_0_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c76e9fe6cae67d11ba62f8b52558b5237ebb2c2c --- /dev/null +++ b/transcoder_all/layer_0_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small/config.json b/transcoder_all/layer_10_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aaab1daf3620b963907cda0e09f50b450e86fca0 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small_affine/config.json b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1c72d1d313944af23cf5c4d8d9c6cc78b6292203 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true, + "type": "transcoder" +} \ 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..ec8c90e5aceded9b9c8c9d50559e77c68c5845be --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 118, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..efdbcc7551d0457a4f9a534608dab3e0eb0d4ba0 --- /dev/null +++ b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 118, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_big_affine/config.json b/transcoder_all/layer_11_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0661d81ffdc1ab6e7a1181ed08e931c9dd3199c8 --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 118, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_small/config.json b/transcoder_all/layer_11_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..15acf77b06959c1c074d69fcdedf0263c3bc3810 --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": false, + "type": "transcoder" +} \ 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..a3329f8a76884de70db37579f7a21ab5125a155c --- /dev/null +++ b/transcoder_all/layer_12_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_big/config.json b/transcoder_all/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b2270df759a07b72d1a3baaa7b9093d40f44ea20 --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_small/config.json b/transcoder_all/layer_13_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b619630ba506ef2849552a1c10c8f61020efe00c --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_big/config.json b/transcoder_all/layer_13_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4d2b2b7f7543f5a6b7444f416ccbfa6a483cc668 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_big_affine/config.json b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..19348c5264d2036dee8d43c9395d3db1dd9e7131 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_small_affine/config.json b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..caa4661139c19292984b975959a5e63f22050d81 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..8d9ae9cfd7add63103e123d21db76c6fb3a6da49 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_small_affine/config.json b/transcoder_all/layer_14_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8243ce90ff16a75472e8b69fe00516952143674f --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_big_affine/config.json b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec597b8e2a40e39dc303ec612247c50ef6623e9f --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_small_affine/config.json b/transcoder_all/layer_14_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..930bed61fa4620014dbbd478b3a414ea345e3eab --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big_affine/config.json b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..227c05b6378f6ab14f26f459464be22b37c1ce1d --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_small_affine/config.json b/transcoder_all/layer_15_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..36fb4d0642c98b2feb424e2f942a472900dc67db --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..122a8fc8cd0595f0e574f1c8e6d13163d7d7fcf4 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_small/config.json b/transcoder_all/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..149c7a89d624c9b53f4ba7f896da23eb2113e1ad --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ 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..ab58b880df7dc8fb59b740de2667ce5be423f2a6 --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big/config.json b/transcoder_all/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fe1aaaa9e9205d989acd29eedcf265b1d26df82a --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_big_affine/config.json b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..04b3f9df1ca9ab33a15c72ad580d8c642cc1983f --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_small/config.json b/transcoder_all/layer_17_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07caaf9ec1dcaa8e7239e77922a18fc84bbf4b8a --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ 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..ed3b1c23553d83c3ba1bb682ef83e5c2522bafbc --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_small_affine/config.json b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aec2da22737a42c86d7cf99b2c74df9b74acfece --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_16k_l0_big/config.json b/transcoder_all/layer_18_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12b762eeb1c216bb94cb3f4c6b0ef92572a1fd69 --- /dev/null +++ b/transcoder_all/layer_18_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ 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..7c545110b0d5c325e0ea5dd748a1126e31e7765e --- /dev/null +++ b/transcoder_all/layer_18_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..e168cb3c09d510f1233479d19ea4b26866194476 --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_big_affine/config.json b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5a8aa380cf29f530d096a3121e15bae456b8abd5 --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ 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..2ccad8550852df08f119763428baf9aaa33ab70b --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_small/config.json b/transcoder_all/layer_19_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7626740d9d0897a54a523e56531eccc6c2887c4e --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_small_affine/config.json b/transcoder_all/layer_19_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7b7ff5500bd97085e8bf60c84cfcd5f5040f88b --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..1bb0376262496302e6a43b4e75db30aa7a0111f5 --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_big/config.json b/transcoder_all/layer_1_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7b6fb268f49a1327c0d241909158c3d988b0a06e --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 65, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_big_affine/config.json b/transcoder_all/layer_1_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b2609d7ba38736c0982038c457c07cad4ba8a9b3 --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 65, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_small_affine/config.json b/transcoder_all/layer_1_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1e53f966633636547e6bbcf96d5fed8cae7aed4f --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_big/config.json b/transcoder_all/layer_20_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b0213f987ce8768416a8a903683b018a8d7531e7 --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_small/config.json b/transcoder_all/layer_20_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..73c48ba1df2813f6495a898f830a2f659b57be5e --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_small_affine/config.json b/transcoder_all/layer_20_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3d061f5d8d355218f5489c5778950d9a668c2bdf --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_small_affine/config.json b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4e01f06c11fed287b65bdb08c761d7387d1556f2 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_big/config.json b/transcoder_all/layer_21_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7c65c303610024b376517920dd4ada206c760f2a --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_16k_l0_big_affine/config.json b/transcoder_all/layer_21_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc44fa4c9d2596150a34d2605f6587f7292e285e --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ 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..8c37a50ba2c00f405b9d4a2600f026e158585643 --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_262k_l0_big_affine/config.json b/transcoder_all/layer_21_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c3cd21edd5ba979a087972456e1ef0660df7be24 --- /dev/null +++ b/transcoder_all/layer_21_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ 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..74142e4235fec8da6ac0e984a68163ee79f35ca6 --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ 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..ab0239d55834e7ed9b670b868db60579a7f3fc1e --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_16k_l0_small/config.json b/transcoder_all/layer_23_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..61be6d261a2c31db9ba46a8521c4e5634340a116 --- /dev/null +++ b/transcoder_all/layer_23_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_small_affine/config.json b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f4633df11736af95e06067f44412a8d660cf1ab --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_big/config.json b/transcoder_all/layer_24_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b85ba01d9c89f1d4aa27383cd59fc7d33ece33a9 --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ 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..7317acc49f60b4307454fe435e9aac03efa5ce75 --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_big/config.json b/transcoder_all/layer_24_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8d1d351b2ee3565d2cb3f8659eb81d19e54108b3 --- /dev/null +++ b/transcoder_all/layer_24_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_small/config.json b/transcoder_all/layer_24_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..63de818e68f4732774388dc49efbd564a8aec48b --- /dev/null +++ b/transcoder_all/layer_24_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_big/config.json b/transcoder_all/layer_25_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6623bca3b74827ba9a5b2bcbd3c261888c0ed14 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_big_affine/config.json b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bbf5c4063bff566904238be048e0ab1708733a85 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ 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..2b8b2e04828c1ec68c424a7131b0b9d3f8b1395f --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_big/config.json b/transcoder_all/layer_26_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b458596d233e37eeb286d38112580ae31084e489 --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_small/config.json b/transcoder_all/layer_26_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8e67ae3c401def46892c6086a089aea9c4a615dc --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_small_affine/config.json b/transcoder_all/layer_26_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a89eea50e058d16491b5251a835cbf6ca0cfc0eb --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_big/config.json b/transcoder_all/layer_26_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..98dadfcb5087ce6c8312a3378991a3a385b83bda --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_big_affine/config.json b/transcoder_all/layer_26_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..707f8e90f7e5e9c92fed6fcb2bfe7028e4c921db --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_small_affine/config.json b/transcoder_all/layer_26_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cae4f80099a293b2fd0297447635563293682681 --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..3207e76d86f7d87fc6311ed0bc1479d3bb0f711f --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_16k_l0_small/config.json b/transcoder_all/layer_28_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a455272cf4931775607a64bfcbd5b3f5f46bd556 --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.28.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.28.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_big/config.json b/transcoder_all/layer_29_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c69e97497f9b0bec98f25831e951c661d9b97be5 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_small/config.json b/transcoder_all/layer_29_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f1017f0f5c2aa899c4b03d397a2424cdd8ad27f6 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_small_affine/config.json b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..21a553792a8cd492436ad6bf6a68dfa4ee99fdbd --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "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": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_small/config.json b/transcoder_all/layer_2_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c771091a6d9eec34bf37762567863bf390c8d93a --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_small_affine/config.json b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a81e0624c2cac1ad13385661027e967d9f52832e --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_big/config.json b/transcoder_all/layer_30_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f29298127293e74df16c442e49dcd8468b73476e --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ 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..6b84ed017f570512d704832d2095516896bbe22c --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_small_affine/config.json b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0aa933c968528463489fc7cf28d69b3c9ed06052 --- /dev/null +++ b/transcoder_all/layer_30_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ 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..c20593447e71ed86753ced39f5d105914a697635 --- /dev/null +++ b/transcoder_all/layer_30_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_262k_l0_small_affine/config.json b/transcoder_all/layer_30_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..21caf6bd89c6dce6bd643d99d13f5fd014a2e2aa --- /dev/null +++ b/transcoder_all/layer_30_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_big_affine/config.json b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..74db9e28457e189a87b3977398ccc4e508ca0908 --- /dev/null +++ b/transcoder_all/layer_31_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_small_affine/config.json b/transcoder_all/layer_31_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bf13decb858dd3e79e87c6297f7a780979913863 --- /dev/null +++ b/transcoder_all/layer_31_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_262k_l0_small/config.json b/transcoder_all/layer_31_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..979ac22dd9393ba20ff943da74d8c1098ced503c --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_262k_l0_small_affine/config.json b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e11cdd6d38c506c6e40b87be5f075d26c54a34dd --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_big/config.json b/transcoder_all/layer_32_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..00088edaf37a916b12256c29dc1c5096f0edac56 --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_big_affine/config.json b/transcoder_all/layer_32_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..13569ac123adfe7c6929eb2680f635e8eb6cdb3d --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_16k_l0_small/config.json b/transcoder_all/layer_32_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9113117513ebcf8b2452763125d52184df34c348 --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ 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..08c3909b0d8012a5778477a25e6b3c0891efa6e7 --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_big/config.json b/transcoder_all/layer_32_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d0457744f34a9557fcc78496715af6aad0a1c5e --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_small/config.json b/transcoder_all/layer_32_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b0d6b01d077f26f8a98e74b79f0a922bf88c4370 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_small_affine/config.json b/transcoder_all/layer_32_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c54feb5cf9e9f34bc9d87f445c956180d4093948 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_small/config.json b/transcoder_all/layer_33_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6e90f4f873129364c81a4d82bf6c8f7fa0c008a1 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_small_affine/config.json b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..354a43c9699c094184e2488c715d6d00812e0fc0 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_big/config.json b/transcoder_all/layer_3_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d7b6f4bac71e46e62d8285da8b64ad274a7270d0 --- /dev/null +++ b/transcoder_all/layer_3_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 75, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small/config.json b/transcoder_all/layer_3_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c4d56a05702787a4ae896e93f8ac27cc1d1521c4 --- /dev/null +++ b/transcoder_all/layer_3_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_big/config.json b/transcoder_all/layer_4_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..38aac01edc8aeeff2615b6130b538af801be03bd --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 81, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aed1e6be98e824aefc61312e07667de47f50758a --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": true, + "type": "transcoder" +} \ 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..44ddf1c438778d83323459fd947496930dd64dcb --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 81, + "affine_connection": false, + "type": "transcoder" +} \ 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..33c7c3ec899bf97ce9cfa53e9b3a53a64fe65b52 --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_small_affine/config.json b/transcoder_all/layer_4_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..584442f7f5b9598bed194ac5065396bfb294ab76 --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_big/config.json b/transcoder_all/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..eed30a2e67ce2088e9d2bd86bbdd547ca736e8f1 --- /dev/null +++ b/transcoder_all/layer_5_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_small_affine/config.json b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c4fae3f0528cdae23ed8bfb2f0224ae2d6567ff7 --- /dev/null +++ b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big/config.json b/transcoder_all/layer_6_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..410e8de2c6044af2e649c95178ebeefd4d33adb3 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 91, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big_affine/config.json b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7b30ce6740da6ea6206b023a3e83f9a5f60bf111 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 91, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_small_affine/config.json b/transcoder_all/layer_6_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b969ac124a21bf9486bc700204f63739b241c065 --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big/config.json b/transcoder_all/layer_7_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a1d808cb7249fea563c366e50377c803df57d06b --- /dev/null +++ b/transcoder_all/layer_7_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dbd81206ff9df87f876f5b0383e21ab2799d0af2 --- /dev/null +++ b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": true, + "type": "transcoder" +} \ 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..fc2eb24d366bed4bbbd6e8d5af24887e4ac925bb --- /dev/null +++ b/transcoder_all/layer_7_width_16k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small_affine/config.json b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..faad0cefb70878c56bee83b7d1dad1db9769b6f0 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_big/config.json b/transcoder_all/layer_8_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7a3b63670fab6e86272fede3a93ec317c25b7380 --- /dev/null +++ b/transcoder_all/layer_8_width_16k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 102, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/config.json b/transcoder_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..36469b8da05389a7324d4e85285f9c71d167abc3 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 102, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big_affine/config.json b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7eb7a963f47f4d3495f8a225ed75a200a73bdca --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 102, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small/config.json b/transcoder_all/layer_8_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c4bce09d991cfe32aaac4e9f99d49b91f113b9a0 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_big_affine/config.json b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ef2efc9e7c8836da7853ee1ffcb4058ca8e4b5bf --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 107, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_small_affine/config.json b/transcoder_all/layer_9_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ab2cbb143d0052dfe5b50d0496fd2fb504ad9236 --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_small_affine/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true, + "type": "transcoder" +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big/config.json b/transcoder_all/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..075e294e81f4c5d6ded91e96a4812d7d8a1cffed --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_big/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 107, + "affine_connection": false, + "type": "transcoder" +} \ 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..ecdeedcdc52ef812cd4e1a8f3a408489e0fcb726 --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,10 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "google/gemma-3-4b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false, + "type": "transcoder" +} \ No newline at end of file