diff --git a/transcoder_all/layer_0_width_16k_l0_big_affine/config.json b/transcoder_all/layer_0_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c672c48fdd4fb52e6db5ec15ffd49a239bb6c38a --- /dev/null +++ b/transcoder_all/layer_0_width_16k_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": 16384, + "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_10_width_16k_l0_small_affine/config.json b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..66059b40696aae92d689fcadd6fab3cd1cb6dd73 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_small/config.json b/transcoder_all/layer_10_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9fd63a8591ecb9cab47caf0bd696336341d3f9ac --- /dev/null +++ b/transcoder_all/layer_10_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_big/config.json b/transcoder_all/layer_11_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..937c4cf5353cf49092140a94ce54a272912d650b --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 101, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_12_width_262k_l0_small_affine/config.json b/transcoder_all/layer_12_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db14713aea2267bf4c60daac1db9e28a7e4517e1 --- /dev/null +++ b/transcoder_all/layer_12_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_small_affine/config.json b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..94add6b78295e6d1bb4c3f7455b2779d91da4161 --- /dev/null +++ b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big/config.json b/transcoder_all/layer_14_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1c29f072e8f14f4850fc265fb2accf7262167ea6 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_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": 16384, + "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_16k_l0_small/config.json b/transcoder_all/layer_14_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c5d2fa4c2e88afeadcc1e6de9f73198e4ddc6234 --- /dev/null +++ b/transcoder_all/layer_14_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_small/config.json b/transcoder_all/layer_14_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f325698283439359b0b1681c2638d21b461a22f0 --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_small/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": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_small_affine/config.json b/transcoder_all/layer_14_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5d8292e8b1712eba733cef0df48fc4f8794385c4 --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_small/config.json b/transcoder_all/layer_17_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e05fe2b9be95a5f72442ee29d2d13c3a40915742 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 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_19_width_16k_l0_big_affine/config.json b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e884b26384fc573e2da746f05310d0741fc71afd --- /dev/null +++ b/transcoder_all/layer_19_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.19.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.19.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_262k_l0_big_affine/config.json b/transcoder_all/layer_19_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..acae7c707812471fe8ed7fe5913eb6e0f0d186db --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_big_affine/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": true +} \ No newline at end of file diff --git a/transcoder_all/layer_19_width_262k_l0_small/config.json b/transcoder_all/layer_19_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..abb4058a7fe9deb8fbc9f52880c4a1f8235eebb7 --- /dev/null +++ b/transcoder_all/layer_19_width_262k_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": 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_19_width_262k_l0_small_affine/config.json b/transcoder_all/layer_19_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c5362e62e0425ba36e99c6e31d04ac723b7afd8 --- /dev/null +++ b/transcoder_all/layer_19_width_262k_l0_small_affine/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": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_big/config.json b/transcoder_all/layer_1_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ba325ae476c1352debe8959655ba24ec317d6653 --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 63, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_16k_l0_small/config.json b/transcoder_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..113e1208e63306d1c201be8da24de6a45559d463 --- /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-it", + "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..b7d17f394a20a8178bf6b2507b0e68e0d9503aee --- /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-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_big_affine/config.json b/transcoder_all/layer_1_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f4d850cd494d65f6b22736516774fce75060f488 --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_big_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": 63, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_20_width_16k_l0_small_affine/config.json b/transcoder_all/layer_20_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e52ca501b35180c89d6842821754231a0281817 --- /dev/null +++ b/transcoder_all/layer_20_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 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_20_width_262k_l0_big/config.json b/transcoder_all/layer_20_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8fd0ffd536362e97300a0a1a4cc603274baef367 --- /dev/null +++ b/transcoder_all/layer_20_width_262k_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": 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_20_width_262k_l0_small_affine/config.json b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12dd98295ed4ce3e31bbbcee26df14235af1fc8c --- /dev/null +++ b/transcoder_all/layer_20_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_21_width_262k_l0_big/config.json b/transcoder_all/layer_21_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f5e5ec11d7c5ba6f5a41444ccb0552528ef15f5f --- /dev/null +++ b/transcoder_all/layer_21_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_22_width_16k_l0_small_affine/config.json b/transcoder_all/layer_22_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c344c3cf91433cd4af5cef5784f01e2eab5c57a --- /dev/null +++ b/transcoder_all/layer_22_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.22.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.22.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_24_width_16k_l0_small/config.json b/transcoder_all/layer_24_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b698d376cb2c9a98db3b25e906d13fe1902c8bc3 --- /dev/null +++ b/transcoder_all/layer_24_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.24.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_25_width_16k_l0_small/config.json b/transcoder_all/layer_25_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d87255071535b22aa8a18caa8ebc7e5be69282df --- /dev/null +++ b/transcoder_all/layer_25_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.25.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.25.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_26_width_16k_l0_small_affine/config.json b/transcoder_all/layer_26_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..64232e7f7d245712190b0bb3a64b3469e2ab869f --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_27_width_16k_l0_small/config.json b/transcoder_all/layer_27_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c052a93d57927edf0b756e6979a217fc8ee79a2 --- /dev/null +++ b/transcoder_all/layer_27_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_29_width_16k_l0_big/config.json b/transcoder_all/layer_29_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9baf4ca813f905fbd5c42ad50f3611f9b4ad759b --- /dev/null +++ b/transcoder_all/layer_29_width_16k_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": 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_29_width_16k_l0_big_affine/config.json b/transcoder_all/layer_29_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3b2bec53cf4720e17f7b80f9e18b068b8b4a4906 --- /dev/null +++ b/transcoder_all/layer_29_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 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_affine/config.json b/transcoder_all/layer_29_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fb21aad18440f0d4c5c1121d5df42d2ebabc6a33 --- /dev/null +++ b/transcoder_all/layer_29_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.29.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.29.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_30_width_16k_l0_small/config.json b/transcoder_all/layer_30_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8f2ea48618189bcfcbe4e52c7e3533306e627743 --- /dev/null +++ b/transcoder_all/layer_30_width_16k_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": 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_30_width_262k_l0_small_affine/config.json b/transcoder_all/layer_30_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e76be9c0f1b5c2ed35534b19f1464b2c95de3020 --- /dev/null +++ b/transcoder_all/layer_30_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.30.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.30.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ 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..639d79136c3efa59629f2b984560f2fe2f842b71 --- /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-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_big_affine/config.json b/transcoder_all/layer_32_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..be8555bc29edc5ac816faa8987e71b5cd3aaa91a --- /dev/null +++ b/transcoder_all/layer_32_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.32.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_33_width_16k_l0_big/config.json b/transcoder_all/layer_33_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b88ea4172c3cc727e142d43efbac4c9b32a8661 --- /dev/null +++ b/transcoder_all/layer_33_width_16k_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": 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_34_width_16k_l0_big/config.json b/transcoder_all/layer_34_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..22f87461470159ada96d7e65a10c6059e48d6b41 --- /dev/null +++ b/transcoder_all/layer_34_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_16k_l0_small_affine/config.json b/transcoder_all/layer_36_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f07960458eddf0aff81d2d4f10671749ca52d992 --- /dev/null +++ b/transcoder_all/layer_36_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_36_width_262k_l0_small_affine/config.json b/transcoder_all/layer_36_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db488d9367278c8efc80670b9372592ee21f5417 --- /dev/null +++ b/transcoder_all/layer_36_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 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_37_width_16k_l0_small_affine/config.json b/transcoder_all/layer_37_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cbdf50c0f8fe289e4c7fd3f99689fee29b110194 --- /dev/null +++ b/transcoder_all/layer_37_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_38_width_262k_l0_small/config.json b/transcoder_all/layer_38_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..33a67e2aacf1c66f222520ffb84d7b6abf553ac4 --- /dev/null +++ b/transcoder_all/layer_38_width_262k_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": 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_39_width_16k_l0_small_affine/config.json b/transcoder_all/layer_39_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..837e236fdee4ab332704c4d38231ec7318bf708e --- /dev/null +++ b/transcoder_all/layer_39_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", + "width": 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_39_width_262k_l0_big/config.json b/transcoder_all/layer_39_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..554cb79bbf7eb52d37369e667e37ad5e2b18afd2 --- /dev/null +++ b/transcoder_all/layer_39_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.39.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.39.post_feedforward_layernorm.output", + "width": 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_3_width_16k_l0_big/config.json b/transcoder_all/layer_3_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ea03fb621cd02eb1f36ce518d6dde0abcccbfb8 --- /dev/null +++ b/transcoder_all/layer_3_width_16k_l0_big/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": false +} \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_small_affine/config.json b/transcoder_all/layer_3_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b23cce48552d45aaddd330759eccba62d0ad450f --- /dev/null +++ b/transcoder_all/layer_3_width_16k_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": 16384, + "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_3_width_262k_l0_big/config.json b/transcoder_all/layer_3_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d408f52dff02e780238ff283dbc5515c499368a5 --- /dev/null +++ b/transcoder_all/layer_3_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 71, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_40_width_262k_l0_small/config.json b/transcoder_all/layer_40_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5905bb10405ae32aafe8dcef8a5170ac93cbd73b --- /dev/null +++ b/transcoder_all/layer_40_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "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..940c127d9ca9837e5732c5b72c8577b10d052150 --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_42_width_262k_l0_small_affine/config.json b/transcoder_all/layer_42_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d55a8c9d0e98d456bf70544c3aca09bbbbacc80d --- /dev/null +++ b/transcoder_all/layer_42_width_262k_l0_small_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": 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_43_width_16k_l0_big_affine/config.json b/transcoder_all/layer_43_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..607fd9f2899608831b3f2ba3070abab627c9c1af --- /dev/null +++ b/transcoder_all/layer_43_width_16k_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": 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_43_width_16k_l0_small_affine/config.json b/transcoder_all/layer_43_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f65a74812184e1c3374c7184bacc0cbe8401a719 --- /dev/null +++ b/transcoder_all/layer_43_width_16k_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": 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_small_affine/config.json b/transcoder_all/layer_44_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7e4124dceaf3cad6be5f7a72631cd9cf021675f6 --- /dev/null +++ b/transcoder_all/layer_44_width_262k_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": 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_45_width_16k_l0_big_affine/config.json b/transcoder_all/layer_45_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fe60240bb01d98751b7aa1aa9794d623227ce7ab --- /dev/null +++ b/transcoder_all/layer_45_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 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_45_width_16k_l0_small/config.json b/transcoder_all/layer_45_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e4f0d62e9421b786a0deaccf3e7e414fdc940088 --- /dev/null +++ b/transcoder_all/layer_45_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_46_width_16k_l0_big/config.json b/transcoder_all/layer_46_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5fbea2fdb2e5e8e47121a12ac61a5195e5df1d2a --- /dev/null +++ b/transcoder_all/layer_46_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.46.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_47_width_262k_l0_big/config.json b/transcoder_all/layer_47_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6cc8b6a189f0f3f02a60bed8dc5658f13da31265 --- /dev/null +++ b/transcoder_all/layer_47_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 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_47_width_262k_l0_small/config.json b/transcoder_all/layer_47_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c42dc1f227e399d788ed6cb7f8745243d7d2663f --- /dev/null +++ b/transcoder_all/layer_47_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_big/config.json b/transcoder_all/layer_4_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6763612de6ea0c7c7fd71f99576ad6ad7a0559bf --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 75, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_big_affine/config.json b/transcoder_all/layer_5_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dbd3dcdadc36f3816dfbb97a6aa7de8600adf584 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_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": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 78, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_big/config.json b/transcoder_all/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f7c9c044e4b93a245215a56f4c37df93a7c09d7e --- /dev/null +++ b/transcoder_all/layer_5_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "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..7b3a2b2d017636964148cd39cc8c32bb9ea86203 --- /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-it", + "architecture": "jump_relu", + "l0": 78, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small/config.json b/transcoder_all/layer_7_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc0b3113d92adb90507da4c38b53a6b612bd670f --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 14, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_big/config.json b/transcoder_all/layer_8_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c467f320a761218d5ecc3a3a5ef1331ae24d3a15 --- /dev/null +++ b/transcoder_all/layer_8_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small/config.json b/transcoder_all/layer_8_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..663448aeef599b51d14400aebd2df2ff9b310ba8 --- /dev/null +++ b/transcoder_all/layer_8_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big/config.json b/transcoder_all/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea456ddc232b204f7cd4f075117d11e4652ea31a --- /dev/null +++ b/transcoder_all/layer_9_width_262k_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": 262144, + "model_name": "gemma-v3-12b-it", + "architecture": "jump_relu", + "l0": 93, + "affine_connection": false +} \ No newline at end of file