diff --git a/transcoder/layer_12_width_16k_l0_big/config.json b/transcoder/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dbb90412a708da6c172fbffd20af47f02478dd9d --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_big/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": 127, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_small/config.json b/transcoder/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..958a247fa5723de6458408b74d1df99151cb20c1 --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_big_affine/config.json b/transcoder/layer_12_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e3174a1ac54417d8c40aff19f165ed24e6483517 --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 127, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_small_affine/config.json b/transcoder/layer_12_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db14713aea2267bf4c60daac1db9e28a7e4517e1 --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_big_affine/config.json b/transcoder/layer_12_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d0a0fbddfd76ba5bb99ea86f672615e050a7fcd --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 127, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_medium_affine/config.json b/transcoder/layer_12_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..00edff2bcb772dddc0066027fa183be0191435d5 --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_medium_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": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 52, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_24_width_16k_l0_medium/config.json b/transcoder/layer_24_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..63539b526b360c263a8a39edbf74c098aaa810f5 --- /dev/null +++ b/transcoder/layer_24_width_16k_l0_medium/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": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_24_width_16k_l0_small/config.json b/transcoder/layer_24_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b698d376cb2c9a98db3b25e906d13fe1902c8bc3 --- /dev/null +++ b/transcoder/layer_24_width_16k_l0_small/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": false +} \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_medium/config.json b/transcoder/layer_24_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f439d63bc72f773e116d81ff7227a296357ee031 --- /dev/null +++ b/transcoder/layer_24_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_medium_affine/config.json b/transcoder/layer_24_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..edb3ccb6cf1978be2b3350e32b17b71622c40447 --- /dev/null +++ b/transcoder/layer_24_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_24_width_262k_l0_small_affine/config.json b/transcoder/layer_24_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bf27a3f1ec79473f301465e037e6b67103ee33c5 --- /dev/null +++ b/transcoder/layer_24_width_262k_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": 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/layer_24_width_65k_l0_big_affine/config.json b/transcoder/layer_24_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dcbbabaa32886de6fff989efd69f4c750719b63f --- /dev/null +++ b/transcoder/layer_24_width_65k_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": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_24_width_65k_l0_small/config.json b/transcoder/layer_24_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a28b0d52f9bea4d30c34b09034af57c12fdbcbdc --- /dev/null +++ b/transcoder/layer_24_width_65k_l0_small/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": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_31_width_16k_l0_big/config.json b/transcoder/layer_31_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a45b8211e6c4b741cef686650bd1e67abc357b78 --- /dev/null +++ b/transcoder/layer_31_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_31_width_16k_l0_small/config.json b/transcoder/layer_31_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f16c144650fb440f97889bd000bb284807a6c71f --- /dev/null +++ b/transcoder/layer_31_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_31_width_65k_l0_medium/config.json b/transcoder/layer_31_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8a8b51d04c03080d9863d54e34b45cf7403b17b8 --- /dev/null +++ b/transcoder/layer_31_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_31_width_65k_l0_medium_affine/config.json b/transcoder/layer_31_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c7c9fb21004802fa21da6a4d9b24fd1771d0cda --- /dev/null +++ b/transcoder/layer_31_width_65k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_41_width_16k_l0_big_affine/config.json b/transcoder/layer_41_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7da75e8cedd703fcc4c8cf61e4b43c64d857a973 --- /dev/null +++ b/transcoder/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": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_41_width_16k_l0_medium/config.json b/transcoder/layer_41_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f946072222e3047f68a5e4591bbe04301ffed55c --- /dev/null +++ b/transcoder/layer_41_width_16k_l0_medium/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": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_41_width_262k_l0_medium_affine/config.json b/transcoder/layer_41_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1262a5158cbff1c24d62150b87ae2f454b2c0371 --- /dev/null +++ b/transcoder/layer_41_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_41_width_65k_l0_big_affine/config.json b/transcoder/layer_41_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6c06b0e44320e0128b99b7fd54f293d0b73fe6a6 --- /dev/null +++ b/transcoder/layer_41_width_65k_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": 65536, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_41_width_65k_l0_small/config.json b/transcoder/layer_41_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..97e692f30ad9b1aa48447a7c8268e3204a562293 --- /dev/null +++ b/transcoder/layer_41_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 65536, + "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_0_width_262k_l0_small_affine/config.json b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..96f9ebc5d13982f624e9c901a226cb0fb6ac549f --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_small_affine/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": true +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_big_affine/config.json b/transcoder_all/layer_10_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..439505d0de875e94f1d465fe8dab7676bc9a6012 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": true +} \ 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..c90fdf0556402dba250faa844c2d9bb81d810126 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_big/config.json b/transcoder_all/layer_10_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..98e038f852f2182c6d3841db39d7572a08267c79 --- /dev/null +++ b/transcoder_all/layer_10_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 97, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_big_affine/config.json b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4af22e8b34d48a4ed58737b04bdae28b8dc20ae2 --- /dev/null +++ b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": true +} \ 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..aebc2ad5b2928b0a305a702412dc04b6481babe5 --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_small_affine/config.json b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e918ccd85db7e763f8c6d21e08b2050c4062b827 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_small_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": 18, + "affine_connection": true +} \ 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..03147664bedc420a8ea2b37f5e5738d4cdd6c682 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_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": 262144, + "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_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..0e1f541fa59881aebc37cf53f13b8ecf737f5525 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_small_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": 18, + "affine_connection": true +} \ 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..cf261fd26cda04554568ab78b99beda410f50e2c --- /dev/null +++ b/transcoder_all/layer_17_width_16k_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": 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_17_width_16k_l0_small/config.json b/transcoder_all/layer_17_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f65356c3f8d5b558cb3cd3ed53b0cb40be45ea2d --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 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_18_width_16k_l0_big/config.json b/transcoder_all/layer_18_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e90902405aa6e14c70c858c78ade65339a4b6820 --- /dev/null +++ b/transcoder_all/layer_18_width_16k_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": 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_18_width_16k_l0_big_affine/config.json b/transcoder_all/layer_18_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f981aba84ba5960b1a8ece3ae6245cd9f1326a81 --- /dev/null +++ b/transcoder_all/layer_18_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 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_1_width_262k_l0_big/config.json b/transcoder_all/layer_1_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec99680478c2abf3123a4d2a7fd3755fe697dbe3 --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_big/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": 63, + "affine_connection": false +} \ 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..79ea404ee2bbc0f3bfa2b0c90b03d7ffedef69e1 --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..1e2bef34ea97d4ababb511d215a3561c7f3045f3 --- /dev/null +++ b/transcoder_all/layer_21_width_16k_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": 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_21_width_16k_l0_small/config.json b/transcoder_all/layer_21_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..477ca3b50dee53321766b9afb066c144b4a6106f --- /dev/null +++ b/transcoder_all/layer_21_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-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_small/config.json b/transcoder_all/layer_21_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91a08deda01b06dec6baa4a6d990d5a2a3890e90 --- /dev/null +++ b/transcoder_all/layer_21_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.21.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.21.post_feedforward_layernorm.output", + "width": 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_262k_l0_small_affine/config.json b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8071ab5956d9fec0df072596ecd6951f32b9ff4c --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_big_affine/config.json b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1982792aaee5083c092c7981c52a97dd79f2c6d0 --- /dev/null +++ b/transcoder_all/layer_24_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..889eb9da69cb5f8658c77d292c884169a99060b5 --- /dev/null +++ b/transcoder_all/layer_26_width_16k_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": 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_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..16eaee1f75fcdd9e30bc1d22a9b5c0f579005fcc --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_big_affine/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": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_big/config.json b/transcoder_all/layer_27_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91d13e225987d0162278619b3214098b46209f23 --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_small_affine/config.json b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c4b3c26fe321897cda3bc847fc617c38f5a64dcc --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_16k_l0_big_affine/config.json b/transcoder_all/layer_28_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d04c8928d5ef8e27c338884a401656ec1cc37832 --- /dev/null +++ b/transcoder_all/layer_28_width_16k_l0_big_affine/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": 120, + "affine_connection": true +} \ 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..ea48da14a8bd255a0469c9c7af160d0d0b0e6df0 --- /dev/null +++ b/transcoder_all/layer_2_width_262k_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": 262144, + "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_262k_l0_big_affine/config.json b/transcoder_all/layer_30_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..eccce2093c23be92cc21b409b1b089e6e923f605 --- /dev/null +++ b/transcoder_all/layer_30_width_262k_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": 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_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..da3d924899b115001940334b8927fa16d1d2106a --- /dev/null +++ b/transcoder_all/layer_31_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_262k_l0_big_affine/config.json b/transcoder_all/layer_31_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fe8a3e943f3662e01a44ff86fc1c3f804d1d236c --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..65b51798eb6908c9ef6314dd67227a6213a9ec71 --- /dev/null +++ b/transcoder_all/layer_31_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "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": "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_big_affine/config.json b/transcoder_all/layer_33_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..80d9ae26ec14980a8d38a247f2cce3d6368d794c --- /dev/null +++ b/transcoder_all/layer_33_width_16k_l0_big_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": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_16k_l0_small/config.json b/transcoder_all/layer_33_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5e43f45b1fabee22c659bc1a88d07c5ac6dafe6f --- /dev/null +++ b/transcoder_all/layer_33_width_16k_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": 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_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..64683ac388c9ae9032472b3192f2881fccc39b07 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-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_small_affine/config.json b/transcoder_all/layer_35_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e435f1b29c301dff0d5e04f292ba34ec74916b5d --- /dev/null +++ b/transcoder_all/layer_35_width_16k_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": 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_35_width_262k_l0_small/config.json b/transcoder_all/layer_35_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fd1025d5d5915f00bafbe726ebd708c42bcc658a --- /dev/null +++ b/transcoder_all/layer_35_width_262k_l0_small/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": false +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_16k_l0_big/config.json b/transcoder_all/layer_36_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1dba1e6e7851526396054edfccf5394a9cce21a8 --- /dev/null +++ b/transcoder_all/layer_36_width_16k_l0_big/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": false +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_262k_l0_small/config.json b/transcoder_all/layer_36_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..110c98f6f723cefb1f09689b95ea5f363bb302b2 --- /dev/null +++ b/transcoder_all/layer_36_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 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_37_width_262k_l0_big/config.json b/transcoder_all/layer_37_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..42f41303762250492d74a9e3acf9693d29070c5e --- /dev/null +++ b/transcoder_all/layer_37_width_262k_l0_big/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": false +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_16k_l0_big_affine/config.json b/transcoder_all/layer_38_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7cb097f2a1e87e9b4aef9133a459e701b9d9cbe4 --- /dev/null +++ b/transcoder_all/layer_38_width_16k_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": 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_39_width_16k_l0_big/config.json b/transcoder_all/layer_39_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2f609143a02a88a67cdde1805d7b69bc44dc82a7 --- /dev/null +++ b/transcoder_all/layer_39_width_16k_l0_big/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_39_width_262k_l0_big_affine/config.json b/transcoder_all/layer_39_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3694c9d92c0052a837638de349b96bb13335a5e0 --- /dev/null +++ b/transcoder_all/layer_39_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", + "width": 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_40_width_262k_l0_big_affine/config.json b/transcoder_all/layer_40_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d18450de6f24947a3ea834ecf200490d7a120c93 --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_41_width_16k_l0_small_affine/config.json b/transcoder_all/layer_41_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cea0ca18d83a9e887e3113c3847a2f9eae3c78c4 --- /dev/null +++ b/transcoder_all/layer_41_width_16k_l0_small_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": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_41_width_262k_l0_small_affine/config.json b/transcoder_all/layer_41_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..150d3196190ff135b503c20b88919cdad227af16 --- /dev/null +++ b/transcoder_all/layer_41_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.41.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_16k_l0_big/config.json b/transcoder_all/layer_42_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5837de4f7b6e9cae5e784cf8a708526855ecbc43 --- /dev/null +++ b/transcoder_all/layer_42_width_16k_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": 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_45_width_16k_l0_big/config.json b/transcoder_all/layer_45_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c9f9121978a12580bd9dacd133606392b13cd4b8 --- /dev/null +++ b/transcoder_all/layer_45_width_16k_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": 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_47_width_16k_l0_big/config.json b/transcoder_all/layer_47_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d928cb75dbb7edbd2c67a23d5d5854fefd2c57a4 --- /dev/null +++ b/transcoder_all/layer_47_width_16k_l0_big/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": 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_4_width_262k_l0_big_affine/config.json b/transcoder_all/layer_4_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea47de3e7fff7816618d5ff399141b5b30a908bb --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 75, + "affine_connection": true +} \ 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..8094251a1860954599c8edf7456203bd95be1e0d --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": true +} \ 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..c174f2fa2fbbb521bbdda633e4dbb08648e8901d --- /dev/null +++ b/transcoder_all/layer_6_width_16k_l0_small_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": 13, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_small_affine/config.json b/transcoder_all/layer_6_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..646c054ab20ce8c4586bd585ca9b4a6b6f37d445 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_l0_small_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": 13, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_small/config.json b/transcoder_all/layer_8_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f6746a1b18e1a332f2c2d96862c42dd4d85cbf2 --- /dev/null +++ b/transcoder_all/layer_8_width_16k_l0_small/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": false +} \ 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..1c68f1bc639e9196e7d385609f49a38c55867124 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": true +} \ No newline at end of file