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..e75010e14b39a062a105a2cc83bf54082e74b3dc --- /dev/null +++ b/transcoder_all/layer_0_width_16k_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": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_small/config.json b/transcoder_all/layer_0_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ef9765471b1ad4d95cfbfe6ec9ecf8a35544e3fa --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_small_affine/config.json b/transcoder_all/layer_11_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b433e691401c2028a07556c044ef5f85842d27fe --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_big/config.json b/transcoder_all/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a9e5714a578bf7f04fb7ef106d38100a91e7c09f --- /dev/null +++ b/transcoder_all/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-pt", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_big_affine/config.json b/transcoder_all/layer_12_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5702506f759bb631998eb53d6553a75948ff02ad --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big/config.json b/transcoder_all/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d2d76ee9bdaf727b1692b22a454f9cb0d08c73d4 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": false +} \ 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..89f494a05d1e13ef15b7cbf651e3d515df3d80c5 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": true +} \ 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..7d95d7a349689ad1af410ba0a2aa0df78cbbac45 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_big/config.json b/transcoder_all/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..119a1e4be123df4017a916d5dfe7be79bd3770de --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_big_affine/config.json b/transcoder_all/layer_15_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c2a33b241c35663e6ac11a5f4384290c83f85176 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small/config.json b/transcoder_all/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3af90bd156dbd3b38d2baf76595b915a99b87df2 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_big/config.json b/transcoder_all/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..11c6ac6bd9c500a0aea126032a02c6915918bba6 --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..c8bff0e097a2c3ce214a8c25405ee873466ec54e --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_16k_l0_small/config.json b/transcoder_all/layer_18_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7523b1b051980ccde0986a5bd62fa7e801c8ec18 --- /dev/null +++ b/transcoder_all/layer_18_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_18_width_262k_l0_small_affine/config.json b/transcoder_all/layer_18_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0843439bd1cee13a9e0ab6f9c868927848cc02bc --- /dev/null +++ b/transcoder_all/layer_18_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_small/config.json b/transcoder_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..52d97054a3bc424927ca26e274263dcb9252bb3a --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_small_affine/config.json b/transcoder_all/layer_1_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0b178523fa9f22d3c626076dc9934b923c804822 --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_small/config.json b/transcoder_all/layer_1_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cffc8753c893767001ec8b509c26397e9491b260 --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ 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..c216a3c6454f1cdbdd016a85d00ae9cc424cb6ed --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ 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..2e6b83182448e2528c1afece7ee9acf00616747b --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_262k_l0_small/config.json b/transcoder_all/layer_20_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d9f4e1149fd420683d95a22bb563bbb8b7ddee96 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "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..650bec21bf6f712fba76905258faf13145409364 --- /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-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_big_affine/config.json b/transcoder_all/layer_22_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..05503e294e9083e0c9c59fc01539c5661d1ef5ab --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..d20cf6fa1b6e810d23c767c5cba1191a66342a59 --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_262k_l0_small/config.json b/transcoder_all/layer_22_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7b19fe7dd3c438b274d1b7cc0424f44284c24e4f --- /dev/null +++ b/transcoder_all/layer_22_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_16k_l0_big/config.json b/transcoder_all/layer_23_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1cf99ab114b5317baedddf04863e96cac20a3229 --- /dev/null +++ b/transcoder_all/layer_23_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_big/config.json b/transcoder_all/layer_23_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3198ffad5fa41d2b2459e864cf20a75c299859a6 --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_big_affine/config.json b/transcoder_all/layer_23_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e47e41a7a4b1e94bbb1b57c407c594479a301bb0 --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_23_width_262k_l0_small/config.json b/transcoder_all/layer_23_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4219a7056a5f474a04fd48ffd360ec3caa9b787c --- /dev/null +++ b/transcoder_all/layer_23_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.23.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..77a51e5a6556334561d95e1626da73f2d4022933 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_small_affine/config.json b/transcoder_all/layer_25_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..34667c53871940c4a50cf593dfdaba3456caeb29 --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_262k_l0_big/config.json b/transcoder_all/layer_25_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..212f816ba288a26cc43c93df6de98f2983590a10 --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_262k_l0_big_affine/config.json b/transcoder_all/layer_25_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4ced300f7783c3c22695d3c8f2fcc156abef154 --- /dev/null +++ b/transcoder_all/layer_25_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ 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..6d81d7ce875038068b5e03a8b1265a1b14122a7d --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_big/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-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..e9320f7fcacb150d649f21578e8484ec25eb4e4a --- /dev/null +++ b/transcoder_all/layer_26_width_16k_l0_small_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": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_small/config.json b/transcoder_all/layer_26_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..27d184e43d551767b981855594534f185f8c34ac --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.26.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.26.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ 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..2f75e006959ae67c39be5600fa9a9f35db94cbca --- /dev/null +++ b/transcoder_all/layer_26_width_262k_l0_small_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-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_16k_l0_big/config.json b/transcoder_all/layer_27_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..96029487c47b5b7c18f6d5458eaa95c144c70081 --- /dev/null +++ b/transcoder_all/layer_27_width_16k_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": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_16k_l0_big_affine/config.json b/transcoder_all/layer_27_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..117116dc96d78fae5ef62b44ad29dc106182c64e --- /dev/null +++ b/transcoder_all/layer_27_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "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_affine/config.json b/transcoder_all/layer_27_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7c7c65efcf2a792862f8b2e1e8b791dca91c67b8 --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_262k_l0_small/config.json b/transcoder_all/layer_27_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..74928a1de1bcb8fc5bc2e320e012df9d68ff7700 --- /dev/null +++ b/transcoder_all/layer_27_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.27.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.27.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_28_width_262k_l0_small_affine/config.json b/transcoder_all/layer_28_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a5d0c2b18d03caadac1a8c346d619404c973744c --- /dev/null +++ b/transcoder_all/layer_28_width_262k_l0_small_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": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ 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..30557623b8b79698341c751c6fb860013f3c46f3 --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 11, + "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..f5495ead254177e9bcc26cc4a4ab01247a108f20 --- /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-pt", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ 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..41fa7be0170fdf6f69ba3364a122a853d049d830 --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_262k_l0_small/config.json b/transcoder_all/layer_30_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e3894ce4a0c8861a2b2e7e55674c53572d180f33 --- /dev/null +++ b/transcoder_all/layer_30_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_big/config.json b/transcoder_all/layer_31_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..560389d5009f6794c96d8ce57ab00d8765db2789 --- /dev/null +++ b/transcoder_all/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-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ 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..a7c4e5f5d77a3438dadf28a355d78b754b75f49d --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_32_width_262k_l0_small/config.json b/transcoder_all/layer_32_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1ffa9009726643e5ac73989a423078719eb8f930 --- /dev/null +++ b/transcoder_all/layer_32_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_262k_l0_big/config.json b/transcoder_all/layer_33_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..669155b23941ced099c3388fe8eacf3a47954654 --- /dev/null +++ b/transcoder_all/layer_33_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_16k_l0_small/config.json b/transcoder_all/layer_34_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ba1f1f43bafaf74ad0a73d435b9aaa2370c07ca4 --- /dev/null +++ b/transcoder_all/layer_34_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_16k_l0_small_affine/config.json b/transcoder_all/layer_34_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1deb160fe8c0e9cf427c6e22b2578a69d4c58f4a --- /dev/null +++ b/transcoder_all/layer_34_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_262k_l0_big/config.json b/transcoder_all/layer_34_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7ba4a2576845bdc7a744cfe11e6d498371df768e --- /dev/null +++ b/transcoder_all/layer_34_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_34_width_262k_l0_big_affine/config.json b/transcoder_all/layer_34_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..444e0e9ca70f81ec13de300a6d6371088cf885f2 --- /dev/null +++ b/transcoder_all/layer_34_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.34.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.34.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_16k_l0_big/config.json b/transcoder_all/layer_35_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6818d81131191a34767f8b2ade24309edecd0c33 --- /dev/null +++ b/transcoder_all/layer_35_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.35.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.35.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_16k_l0_small/config.json b/transcoder_all/layer_35_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0333bf353bf2c81d8e2add3555f8f34a5a200d1f --- /dev/null +++ b/transcoder_all/layer_35_width_16k_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": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_262k_l0_small_affine/config.json b/transcoder_all/layer_35_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c332eef2782f5ed860808daa657c1d65e611d0ea --- /dev/null +++ b/transcoder_all/layer_35_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.35.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.35.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_16k_l0_small/config.json b/transcoder_all/layer_36_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..844c990c65b855a03c93f635dc9e243046ff00f9 --- /dev/null +++ b/transcoder_all/layer_36_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_16k_l0_big_affine/config.json b/transcoder_all/layer_37_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..54fc2a67aa35af2bee80deeb82174c51d65317fb --- /dev/null +++ b/transcoder_all/layer_37_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_262k_l0_big_affine/config.json b/transcoder_all/layer_37_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df20cebd16c7924004e53cbbb94b1e6d9208ff2a --- /dev/null +++ b/transcoder_all/layer_37_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_37_width_262k_l0_small_affine/config.json b/transcoder_all/layer_37_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9373d7adcacd05cf68da2f5d23b6718d57ad72a8 --- /dev/null +++ b/transcoder_all/layer_37_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.37.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_262k_l0_big/config.json b/transcoder_all/layer_38_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ac304d3e0373910cc390f36fe2883e93b495d765 --- /dev/null +++ b/transcoder_all/layer_38_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "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..0d7cf8a719d56c3b2057c36e1b71285234a3ef7f --- /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-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small_affine/config.json b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0dd29e2a5b98259dad26ed3b8aa425fdfd8a5e5b --- /dev/null +++ b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_16k_l0_big_affine/config.json b/transcoder_all/layer_40_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..de25bd822decd3b641e291445be46ef36dbf1df9 --- /dev/null +++ b/transcoder_all/layer_40_width_16k_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": 16384, + "model_name": "gemma-v3-12b-pt", + "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/config.json b/transcoder_all/layer_40_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dd0b96a129db186298b7375f58a4bc47ac06c9c3 --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_262k_l0_small_affine/config.json b/transcoder_all/layer_40_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d543aee242da13cf17365f3df2749ff246a33eb5 --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "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/config.json b/transcoder_all/layer_41_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f2bd31b3169bb60d143791149dcab392f71d36ae --- /dev/null +++ b/transcoder_all/layer_41_width_262k_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": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_16k_l0_small/config.json b/transcoder_all/layer_43_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fc9e4b4fd8fd70b93554d71fc1c2109722ae071d --- /dev/null +++ b/transcoder_all/layer_43_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_big/config.json b/transcoder_all/layer_43_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c1687d94bcf11898ac253ad1454653c9df76ab1b --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_small/config.json b/transcoder_all/layer_43_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c95e02e02b15d2bf71243b9bb7f194b2d6f1270 --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_small_affine/config.json b/transcoder_all/layer_43_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0a2348266b918d41c6d25e5bc278c67e49e389f --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_small/config.json b/transcoder_all/layer_44_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5e23bafc0933a7dc0ad39e0617335e87d0829948 --- /dev/null +++ b/transcoder_all/layer_44_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_262k_l0_big/config.json b/transcoder_all/layer_44_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fbd079af92fdffddb3ac3f2b47e21322f02a3c29 --- /dev/null +++ b/transcoder_all/layer_44_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_45_width_262k_l0_small_affine/config.json b/transcoder_all/layer_45_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..03d725fb9da7772b42bcb062953857ce5b2ada5a --- /dev/null +++ b/transcoder_all/layer_45_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_46_width_262k_l0_big_affine/config.json b/transcoder_all/layer_46_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a331be3ff6bdd0410c795272598eae9f93fed337 --- /dev/null +++ b/transcoder_all/layer_46_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_47_width_16k_l0_small_affine/config.json b/transcoder_all/layer_47_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec58cfc93471720910467fd9ab712823d4059726 --- /dev/null +++ b/transcoder_all/layer_47_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fe439b10438ba0b4520fe515d9d1eb664e33c717 --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": true +} \ 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..2d904cb88ba86dfbd24158e96ff9120fe34e4a94 --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_big/config.json b/transcoder_all/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c4b54a5311691da02a8c51df3b9c29cfdb6e2f3 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 78, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_big_affine/config.json b/transcoder_all/layer_5_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c66ba9c54531e44e1ad5dbcde0e1e91b39b19d72 --- /dev/null +++ b/transcoder_all/layer_5_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 78, + "affine_connection": true +} \ 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..94f75a484568c7c547cae77f478b240033888dba --- /dev/null +++ b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 13, + "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..b28efabd9f81148fbd6cae959e47f06e98950b22 --- /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-pt", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_big_affine/config.json b/transcoder_all/layer_6_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df205e898c027cf0ca2e5f094302053a7cdb88d6 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 82, + "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..03426ca39b289a222670d4e1a295e0b12461e817 --- /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-pt", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_big/config.json b/transcoder_all/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..71846c8ac568742cb6ad3f3ad4e047d89d653fb4 --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 93, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_small/config.json b/transcoder_all/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f5dec0e401f2d434555133d4772fde2ed51bf45d --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-pt", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file