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..f1d6d2fb58ace720211e792b7ccdf95537a9c83c --- /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-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_big_affine/config.json b/transcoder_all/layer_0_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e14b1fd527b14b9ffa7ec1bfe6c9ad7e734f5b03 --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_big_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": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_big_affine/config.json b/transcoder_all/layer_11_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f07f08d3587bf5a29b4de5c95300613abaf66c7d --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_big_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-it", + "architecture": "jump_relu", + "l0": 101, + "affine_connection": true +} \ 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..021eefab15880eaa2d6ffddbaac9007974770e25 --- /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-it", + "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..53280e665b4504e3308132f803cb9044c9666de7 --- /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-it", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_small/config.json b/transcoder_all/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..958a247fa5723de6458408b74d1df99151cb20c1 --- /dev/null +++ b/transcoder_all/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_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..45021fc97d58429dbc46f4999d6cafb0c31c22bc --- /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-it", + "architecture": "jump_relu", + "l0": 105, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_big/config.json b/transcoder_all/layer_14_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cb3228c7d8e0b92efb97c0fc30068681d18ec6ac --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_big/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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 112, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_big_affine/config.json b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bbb13f38d180679b44d9aa069419abed530307f1 --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 112, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big_affine/config.json b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9ce4c676348a3ba9ed085e151e23ea1aa83220da --- /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-it", + "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..3aeebd1c5841826852be532e9fc480c35c760b79 --- /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-it", + "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..8a2098460e356475ff48350538153de9566121db --- /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-it", + "architecture": "jump_relu", + "l0": 116, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_16k_l0_big/config.json b/transcoder_all/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b74dc186f1ceafe2ebf303aa7f7d500feb1c61fe --- /dev/null +++ b/transcoder_all/layer_16_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big/config.json b/transcoder_all/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2b1862cbe77240a600bf54066b7b1d419dae7dcb --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e7b0b37424c3d9a858dc324db8042c10a7c98dd7 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small_affine/config.json b/transcoder_all/layer_16_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9739d12ce83318a5a538a8b6bc4820b99c3779d3 --- /dev/null +++ b/transcoder_all/layer_16_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 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_17_width_16k_l0_big/config.json b/transcoder_all/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e6c583388677a7330a1d017348c8622642128d1f --- /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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_small_affine/config.json b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..afe38fbcd8ddf706fb47bc6e136956aa54dd754c --- /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-it", + "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..af3ba9f72b1a64ce86a32a7effd489806b728ab3 --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_16k_l0_small/config.json b/transcoder_all/layer_19_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..299c52c56e13f998571853fcb5dc57d42806ea97 --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 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_19_width_262k_l0_big/config.json b/transcoder_all/layer_19_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..387e61a97c1d781bb22f129ab5984b570abdb1b6 --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 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_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..ad50255bd0675037bf55175ca921c97367f27205 --- /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-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_262k_l0_small_affine/config.json b/transcoder_all/layer_24_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bf27a3f1ec79473f301465e037e6b67103ee33c5 --- /dev/null +++ b/transcoder_all/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_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..4d9d071d8bad1f83f01e8f4b5c2b092b7da2ddcb --- /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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_262k_l0_big/config.json b/transcoder_all/layer_26_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a0c8e84a38aa09777a546bc783e34e214604652 --- /dev/null +++ b/transcoder_all/layer_26_width_262k_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": 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_16k_l0_big/config.json b/transcoder_all/layer_27_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fb8f9bf6a223a499015ab365d4800e286992bd50 --- /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-it", + "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..e12218fc9fa0e483cd66fcae4b416a7f1d051753 --- /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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_262k_l0_big/config.json b/transcoder_all/layer_29_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..56ec942960e424c620580f68b74a04a581516060 --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_big_affine/config.json b/transcoder_all/layer_2_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2bd4296df68095d5a86c5bba8585a222b28aef30 --- /dev/null +++ b/transcoder_all/layer_2_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 67, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_big/config.json b/transcoder_all/layer_2_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9faa5ac140affb6ac20577f97a8c947a980543dc --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 67, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_small_affine/config.json b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b81feb6a8706b578fc1f10401be23b52108e5a98 --- /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-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_31_width_16k_l0_small/config.json b/transcoder_all/layer_31_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f16c144650fb440f97889bd000bb284807a6c71f --- /dev/null +++ b/transcoder_all/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_all/layer_32_width_16k_l0_big/config.json b/transcoder_all/layer_32_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c56c18d061465906f41f0fa30d1e54185694c10a --- /dev/null +++ b/transcoder_all/layer_32_width_16k_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": 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_32_width_16k_l0_small/config.json b/transcoder_all/layer_32_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cf4add3b2bbf3f158e782e531e63faa35805d785 --- /dev/null +++ b/transcoder_all/layer_32_width_16k_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": 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_32_width_262k_l0_big/config.json b/transcoder_all/layer_32_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c04f540749d3f4d79d819374a5a4b3e81b9733b5 --- /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-it", + "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..19a0c6f93d29117c1a8f7d963b91f1e4a502c666 --- /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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_35_width_262k_l0_big/config.json b/transcoder_all/layer_35_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fa1e5e727da15a6f767897b088a0b6f7b4fdc686 --- /dev/null +++ b/transcoder_all/layer_35_width_262k_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": 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_36_width_16k_l0_small/config.json b/transcoder_all/layer_36_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..26c716c180a6bed44e7543cf18cfcee16b7f6e56 --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_262k_l0_big/config.json b/transcoder_all/layer_36_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5b58bc15714df1e04f543cd99bf8a382b0a51d6a --- /dev/null +++ b/transcoder_all/layer_36_width_262k_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": 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_37_width_16k_l0_big/config.json b/transcoder_all/layer_37_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c456892dbbab3cdb80e98c79022b32bf1bf2b243 --- /dev/null +++ b/transcoder_all/layer_37_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_16k_l0_small/config.json b/transcoder_all/layer_38_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6516d8787c5c645d8613b2eadffb5cb7de3ac121 --- /dev/null +++ b/transcoder_all/layer_38_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_big_affine/config.json b/transcoder_all/layer_3_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..97df36539529fd9d33a6bce651b1c85a8f788d55 --- /dev/null +++ b/transcoder_all/layer_3_width_16k_l0_big_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 71, + "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..3375fe2f9954b166db1e281cfd5ccdac6a74d5f6 --- /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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_16k_l0_big_affine/config.json b/transcoder_all/layer_42_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a8a6d1bd32a2dea9fd723a7fb6b9e11f2cce7102 --- /dev/null +++ b/transcoder_all/layer_42_width_16k_l0_big_affine/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": true +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_262k_l0_small/config.json b/transcoder_all/layer_42_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b3e047743ac5d2081beceaf20f486b74ac5e960f --- /dev/null +++ b/transcoder_all/layer_42_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.42.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.42.post_feedforward_layernorm.output", + "width": 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_43_width_262k_l0_big_affine/config.json b/transcoder_all/layer_43_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..85cd6c999e96a710ac6f369076eddaf666615176 --- /dev/null +++ b/transcoder_all/layer_43_width_262k_l0_big_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-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_43_width_262k_l0_small/config.json b/transcoder_all/layer_43_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b98f45004b4c18822a5708ce78831170db7c7e85 --- /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-it", + "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..39cf3b9bcd40965cad0024802d9954dc71176363 --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_small/config.json b/transcoder_all/layer_44_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6810524b27e38b11b8ae15ab8fd4787835f31fe --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_16k_l0_small_affine/config.json b/transcoder_all/layer_44_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..88425e6e902cb561e5d8e06ce8f7e90dd05d48c7 --- /dev/null +++ b/transcoder_all/layer_44_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_44_width_262k_l0_big_affine/config.json b/transcoder_all/layer_44_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9ec653572602af8924e7d10852eb5cb78bf1f7d5 --- /dev/null +++ b/transcoder_all/layer_44_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.44.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.44.post_feedforward_layernorm.output", + "width": 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_46_width_16k_l0_big_affine/config.json b/transcoder_all/layer_46_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..85befdb0e59de69025039adc9533b0df1594ec45 --- /dev/null +++ b/transcoder_all/layer_46_width_16k_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": 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_46_width_262k_l0_small_affine/config.json b/transcoder_all/layer_46_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0cbf04189c26f6b7896664e5642b74743515b845 --- /dev/null +++ b/transcoder_all/layer_46_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "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..171eb75bda2106fbaa77631803984b3f3bf6d038 --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small/config.json b/transcoder_all/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f27ce0d55d9c012737151a714247b1b16d433e9f --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big/config.json b/transcoder_all/layer_7_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..106d04cbc3c30fa4b7cd04fdaa48cc334e0e2f3d --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small_affine/config.json b/transcoder_all/layer_8_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..98207fddb72f92d4d2b20392478578eed737ebba --- /dev/null +++ b/transcoder_all/layer_8_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": true +} \ No newline at end of file