diff --git a/clt/width_262k_l0_big/config.json b/clt/width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e862c095410cefc1790f28b33c4e9d96ca9c92af --- /dev/null +++ b/clt/width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output", + "width": 262080, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_524k_l0_medium_affine/config.json b/clt/width_524k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e3f4e3a7b7697f90ab765d3e6c8fae418551f89c --- /dev/null +++ b/clt/width_524k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output", + "width": 524160, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": true +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_1m_l0_big/config.json b/crosscoder/layer_5_9_12_15_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd372a687f86182e83e4f4654273b80ee0073aa1 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_1m_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 1048576, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_1m_l0_medium/config.json b/crosscoder/layer_5_9_12_15_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..74214293c55830e0980874c49f0af75338a64076 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_1m_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 1048576, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_262k_l0_big/config.json b/crosscoder/layer_5_9_12_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2990ced83e5aff935d511a4d00b0b8ce3497ba2a --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_262k_l0_medium/config.json b/crosscoder/layer_5_9_12_15_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..603995d44766ab00edfffda793ea3c313a3ce8bf --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_524k_l0_big/config.json b/crosscoder/layer_5_9_12_15_width_524k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..744dab41d50c6b1177b9fb92c4dc4b1a1f8b6bb0 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_524k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 524288, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_524k_l0_medium/config.json b/crosscoder/layer_5_9_12_15_width_524k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..64f8de19e183a31f7d8611f1d9d92517ad12d878 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_524k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 524288, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_65k_l0_big/config.json b/crosscoder/layer_5_9_12_15_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ae56c86fac540f17a7b43f2d32ab3abdbb2ea9f3 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/crosscoder/layer_5_9_12_15_width_65k_l0_medium/config.json b/crosscoder/layer_5_9_12_15_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39ad26fbb87e6dc0b7dd8c589c06ad0a3379de77 --- /dev/null +++ b/crosscoder/layer_5_9_12_15_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "hf_hook_point_out": "model.layers.{5.0,9.5,12.200000000000001,15.799999999999999}.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_16k_l0_big/config.json b/mlp_out/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3b9091c36fc7dfdae34f2c876c826e4fa156894e --- /dev/null +++ b/mlp_out/layer_12_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_16k_l0_medium/config.json b/mlp_out/layer_12_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..857266cea9fe0daabd37c6e440af4cbd41e179e9 --- /dev/null +++ b/mlp_out/layer_12_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_16k_l0_small/config.json b/mlp_out/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4699a11ca5ed92998469629a6bdb69f91f215bf1 --- /dev/null +++ b/mlp_out/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_262k_l0_big/config.json b/mlp_out/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4776e897c8f771e2df5a5e0e130914d41737415e --- /dev/null +++ b/mlp_out/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_262k_l0_medium/config.json b/mlp_out/layer_12_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0889b9b8b849b17960a6e13b89c82095d2fc669e --- /dev/null +++ b/mlp_out/layer_12_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_262k_l0_small/config.json b/mlp_out/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84f463d615920c1161012b4f8ec0ab75e19fc4ae --- /dev/null +++ b/mlp_out/layer_12_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_65k_l0_big/config.json b/mlp_out/layer_12_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c7a7e91805dba60eb7cf73630dfe5ce42094322 --- /dev/null +++ b/mlp_out/layer_12_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_65k_l0_medium/config.json b/mlp_out/layer_12_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..952a280a72a69dd316c37c3e75efe41abeadcf3f --- /dev/null +++ b/mlp_out/layer_12_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_12_width_65k_l0_small/config.json b/mlp_out/layer_12_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1bfcfaca8c02a9ef68df43f79df689be46fcf74c --- /dev/null +++ b/mlp_out/layer_12_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_16k_l0_big/config.json b/mlp_out/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e268ee25220991c301d09756e30f6ca0e156c364 --- /dev/null +++ b/mlp_out/layer_15_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_16k_l0_medium/config.json b/mlp_out/layer_15_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f9fc520dbd17bee4d09d2bf21a1553791b192b84 --- /dev/null +++ b/mlp_out/layer_15_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_16k_l0_small/config.json b/mlp_out/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b31fdcd386bfeec459039b7ee038cb9613c133ff --- /dev/null +++ b/mlp_out/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_262k_l0_big/config.json b/mlp_out/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..655420d8987e2e63df6fbe941191324fb203754a --- /dev/null +++ b/mlp_out/layer_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_262k_l0_medium/config.json b/mlp_out/layer_15_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9237d43724675f0017238af132d7593925b440f3 --- /dev/null +++ b/mlp_out/layer_15_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_262k_l0_small/config.json b/mlp_out/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bdf4cb59b767b38f2e8f218c09a409e7919975c5 --- /dev/null +++ b/mlp_out/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_65k_l0_big/config.json b/mlp_out/layer_15_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..65a06e05ac40284c0dcb270ec0328f1b1f0e2b30 --- /dev/null +++ b/mlp_out/layer_15_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_65k_l0_medium/config.json b/mlp_out/layer_15_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a7de5289c642d4a25b14f4bc8661aaa4c9b211c --- /dev/null +++ b/mlp_out/layer_15_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_15_width_65k_l0_small/config.json b/mlp_out/layer_15_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..061a70c76b07f14c7b00ba08f34745827eafb3a5 --- /dev/null +++ b/mlp_out/layer_15_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_16k_l0_big/config.json b/mlp_out/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9c94ba6c9fc9ef8a4e36c07beda6b73b04345710 --- /dev/null +++ b/mlp_out/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_16k_l0_medium/config.json b/mlp_out/layer_5_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4713569560396c734c7534e22076569e952fef9 --- /dev/null +++ b/mlp_out/layer_5_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_16k_l0_small/config.json b/mlp_out/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..30b468474c5f93e9ba772fa4fbf3dab139ef0e28 --- /dev/null +++ b/mlp_out/layer_5_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_262k_l0_big/config.json b/mlp_out/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ec8f6fa2b4f409ffb3e4b4f72b3d1ed7c6c7dff --- /dev/null +++ b/mlp_out/layer_5_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_262k_l0_medium/config.json b/mlp_out/layer_5_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4dcbfd0081248a70058d2bb16b49cef4a556bfa --- /dev/null +++ b/mlp_out/layer_5_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_262k_l0_small/config.json b/mlp_out/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c6d871f7619bcb665c395af556cc19c756025dfa --- /dev/null +++ b/mlp_out/layer_5_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_65k_l0_big/config.json b/mlp_out/layer_5_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e0d754dc66803add301deeea3ee160c63e5b9df --- /dev/null +++ b/mlp_out/layer_5_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_65k_l0_medium/config.json b/mlp_out/layer_5_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..68c5347eae2308304e4f5a437f0ab0c01a7a4a1a --- /dev/null +++ b/mlp_out/layer_5_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_5_width_65k_l0_small/config.json b/mlp_out/layer_5_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..02808e14df3311da01e7d210aaf1a16a7bfa22c9 --- /dev/null +++ b/mlp_out/layer_5_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_16k_l0_big/config.json b/mlp_out/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b1f9d941927ac68a906653355d25027777007ef0 --- /dev/null +++ b/mlp_out/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_16k_l0_medium/config.json b/mlp_out/layer_9_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9c46476d5b4bb0970c1d1ae3863a785eeb318c99 --- /dev/null +++ b/mlp_out/layer_9_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_16k_l0_small/config.json b/mlp_out/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39288522b0ffab750eb98e79b80f57284e5c95a5 --- /dev/null +++ b/mlp_out/layer_9_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_262k_l0_big/config.json b/mlp_out/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..78b6dc8a9a4954d5ddbf31b3276c6db990b1d37a --- /dev/null +++ b/mlp_out/layer_9_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_262k_l0_medium/config.json b/mlp_out/layer_9_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c3166f82716b9926eef5d77f39679046d44d83ef --- /dev/null +++ b/mlp_out/layer_9_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_262k_l0_small/config.json b/mlp_out/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7fce4e84f82961fd495e31fdaf1f75f8c2541383 --- /dev/null +++ b/mlp_out/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_65k_l0_big/config.json b/mlp_out/layer_9_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4b57702e2878ca69688cff5dc5a26bae470489d4 --- /dev/null +++ b/mlp_out/layer_9_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_65k_l0_medium/config.json b/mlp_out/layer_9_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..29e9f2914fe9d3c8e0d552751a0917f9b90bcc11 --- /dev/null +++ b/mlp_out/layer_9_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_9_width_65k_l0_small/config.json b/mlp_out/layer_9_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bad332136a18b8ada5f5f6f3479523311b8cba8f --- /dev/null +++ b/mlp_out/layer_9_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_0_width_16k_l0_big/config.json b/mlp_out_all/layer_0_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..38e85c69e044739a540f935d9e5d6abaf9aed390 --- /dev/null +++ b/mlp_out_all/layer_0_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_0_width_16k_l0_small/config.json b/mlp_out_all/layer_0_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..925d4cb258fd9e731115f7eb806f5a9dc78593bc --- /dev/null +++ b/mlp_out_all/layer_0_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_0_width_262k_l0_big/config.json b/mlp_out_all/layer_0_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db4ea14c707ba6deccae807a994e772e9e5100a2 --- /dev/null +++ b/mlp_out_all/layer_0_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_0_width_262k_l0_small/config.json b/mlp_out_all/layer_0_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..21124c6f8b1c93fe71bb5d0d9a0be630fba80455 --- /dev/null +++ b/mlp_out_all/layer_0_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_10_width_16k_l0_big/config.json b/mlp_out_all/layer_10_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..01d79dcb232b4d79bd06e03e7bdeeab9ad30f16f --- /dev/null +++ b/mlp_out_all/layer_10_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_10_width_16k_l0_small/config.json b/mlp_out_all/layer_10_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c43064b216320b98f51346c6e95377e3dc493195 --- /dev/null +++ b/mlp_out_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_10_width_262k_l0_big/config.json b/mlp_out_all/layer_10_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0391add096a70e40c0f99db6d85bdc6969037b36 --- /dev/null +++ b/mlp_out_all/layer_10_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_10_width_262k_l0_small/config.json b/mlp_out_all/layer_10_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7b1164b99437bd8140ca4bc23651a621030dfe14 --- /dev/null +++ b/mlp_out_all/layer_10_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_11_width_16k_l0_big/config.json b/mlp_out_all/layer_11_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1a13bea3581ca9ed8c09284ee6b7ecf7ffe30edf --- /dev/null +++ b/mlp_out_all/layer_11_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_11_width_16k_l0_small/config.json b/mlp_out_all/layer_11_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aae78685d9e0213b21ac8b35955bec2da87f8d02 --- /dev/null +++ b/mlp_out_all/layer_11_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_11_width_262k_l0_big/config.json b/mlp_out_all/layer_11_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2705921fff3ddbb7bdd46dfdb0e48e9729372d0b --- /dev/null +++ b/mlp_out_all/layer_11_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_11_width_262k_l0_small/config.json b/mlp_out_all/layer_11_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a701864c3b821a4ea4f2e5e4e1667c071f5f2826 --- /dev/null +++ b/mlp_out_all/layer_11_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_12_width_16k_l0_big/config.json b/mlp_out_all/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bbd6aec50dda2a398d523494cfff76ae8c3b9edf --- /dev/null +++ b/mlp_out_all/layer_12_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_12_width_16k_l0_small/config.json b/mlp_out_all/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4699a11ca5ed92998469629a6bdb69f91f215bf1 --- /dev/null +++ b/mlp_out_all/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_12_width_262k_l0_big/config.json b/mlp_out_all/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bec572c190f3a1ce1293a2f2173585a811c478d5 --- /dev/null +++ b/mlp_out_all/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_12_width_262k_l0_small/config.json b/mlp_out_all/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84f463d615920c1161012b4f8ec0ab75e19fc4ae --- /dev/null +++ b/mlp_out_all/layer_12_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_13_width_16k_l0_big/config.json b/mlp_out_all/layer_13_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ac80cc0961fa3072fc05ae5eb5bbb958e1010dd0 --- /dev/null +++ b/mlp_out_all/layer_13_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_13_width_16k_l0_small/config.json b/mlp_out_all/layer_13_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..09e80732b55b03f0695f3a778951bf967cd98a3e --- /dev/null +++ b/mlp_out_all/layer_13_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_13_width_262k_l0_big/config.json b/mlp_out_all/layer_13_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..96b2cb3a4645534767aeea69134315eb9fe2b043 --- /dev/null +++ b/mlp_out_all/layer_13_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_13_width_262k_l0_small/config.json b/mlp_out_all/layer_13_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39de3c33608ffdbe51fcb97e7a307838f4d5e3b5 --- /dev/null +++ b/mlp_out_all/layer_13_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_14_width_16k_l0_big/config.json b/mlp_out_all/layer_14_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a227d4d71878e175f38ac3d85343b0e5954c8cb7 --- /dev/null +++ b/mlp_out_all/layer_14_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_14_width_16k_l0_small/config.json b/mlp_out_all/layer_14_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..456ef83f857ad98c39d6184c264b86b25d9d95a9 --- /dev/null +++ b/mlp_out_all/layer_14_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_14_width_262k_l0_big/config.json b/mlp_out_all/layer_14_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1bf269ef33c5a72356613bcc26ee6fd226415af2 --- /dev/null +++ b/mlp_out_all/layer_14_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_14_width_262k_l0_small/config.json b/mlp_out_all/layer_14_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7a96c0cd949fd3a354329bee9b988117c03b268e --- /dev/null +++ b/mlp_out_all/layer_14_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_15_width_16k_l0_big/config.json b/mlp_out_all/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2f22b4348c12c8a76557b6cde93fc3ff6a1de2e5 --- /dev/null +++ b/mlp_out_all/layer_15_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_15_width_16k_l0_small/config.json b/mlp_out_all/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b31fdcd386bfeec459039b7ee038cb9613c133ff --- /dev/null +++ b/mlp_out_all/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_15_width_262k_l0_big/config.json b/mlp_out_all/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b936ebc8ca3b3b2697c0bb7d232988aaaf75372 --- /dev/null +++ b/mlp_out_all/layer_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_15_width_262k_l0_small/config.json b/mlp_out_all/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bdf4cb59b767b38f2e8f218c09a409e7919975c5 --- /dev/null +++ b/mlp_out_all/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_16_width_16k_l0_big/config.json b/mlp_out_all/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6c301b1c9f55208e7f1c19eb98c0dc5c59ce9611 --- /dev/null +++ b/mlp_out_all/layer_16_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_16_width_16k_l0_small/config.json b/mlp_out_all/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a1b3fd4552005695a5351c49c17a83c78d81520 --- /dev/null +++ b/mlp_out_all/layer_16_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_16_width_262k_l0_big/config.json b/mlp_out_all/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..405fea3dfe8430b9473e332cc97ecab190179e75 --- /dev/null +++ b/mlp_out_all/layer_16_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_16_width_262k_l0_small/config.json b/mlp_out_all/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5bda8d034bbd55cbbeb9a80677b0598844108d66 --- /dev/null +++ b/mlp_out_all/layer_16_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_17_width_16k_l0_big/config.json b/mlp_out_all/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6cfe1f984da35e3f5fcebc2d4657c5a3ca02115b --- /dev/null +++ b/mlp_out_all/layer_17_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_17_width_16k_l0_small/config.json b/mlp_out_all/layer_17_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..98976eabd2f05a598e6f4cea6495553df7971d77 --- /dev/null +++ b/mlp_out_all/layer_17_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_17_width_262k_l0_big/config.json b/mlp_out_all/layer_17_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6a25cdbd3ffdc89b9d37f77ab56cfcdd3f966471 --- /dev/null +++ b/mlp_out_all/layer_17_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_17_width_262k_l0_small/config.json b/mlp_out_all/layer_17_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5302c447211a49d3bdb9658c142ddcac3f2daa90 --- /dev/null +++ b/mlp_out_all/layer_17_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_1_width_16k_l0_big/config.json b/mlp_out_all/layer_1_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a45d561d5cbb75b918a3d9daaf019cf2a79692ab --- /dev/null +++ b/mlp_out_all/layer_1_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_1_width_16k_l0_small/config.json b/mlp_out_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4dfee370211c8b3ec001386207a4935d7d60a6f3 --- /dev/null +++ b/mlp_out_all/layer_1_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_1_width_262k_l0_big/config.json b/mlp_out_all/layer_1_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..72170488877067113a44c6ff4d99bffac505c4d5 --- /dev/null +++ b/mlp_out_all/layer_1_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_1_width_262k_l0_small/config.json b/mlp_out_all/layer_1_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..55c839e55fc92b092c0fc1e2c69e5993f238ebad --- /dev/null +++ b/mlp_out_all/layer_1_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_2_width_16k_l0_big/config.json b/mlp_out_all/layer_2_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..354afe925f7315e4c66620cd2a724329da0e3388 --- /dev/null +++ b/mlp_out_all/layer_2_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_2_width_16k_l0_small/config.json b/mlp_out_all/layer_2_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a1eadabf1fd49ad8a4c7eb63785bdd270010dd13 --- /dev/null +++ b/mlp_out_all/layer_2_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_2_width_262k_l0_big/config.json b/mlp_out_all/layer_2_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c6215800b3e60b93bda9f194b9ebafc3831193b --- /dev/null +++ b/mlp_out_all/layer_2_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_2_width_262k_l0_small/config.json b/mlp_out_all/layer_2_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2ad96c83bd032251588c8d0c407a9c1f81a938a9 --- /dev/null +++ b/mlp_out_all/layer_2_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_3_width_16k_l0_big/config.json b/mlp_out_all/layer_3_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b12ae3343df7d3af7486ea346a7fc327a83d8cf2 --- /dev/null +++ b/mlp_out_all/layer_3_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_3_width_16k_l0_small/config.json b/mlp_out_all/layer_3_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6309099498903c46ff7b6c3b14baa61b8b77c6bf --- /dev/null +++ b/mlp_out_all/layer_3_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_3_width_262k_l0_big/config.json b/mlp_out_all/layer_3_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a3a14408f3115c20007e6d300320c051058c39cd --- /dev/null +++ b/mlp_out_all/layer_3_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_3_width_262k_l0_small/config.json b/mlp_out_all/layer_3_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4e4a6b65907ac9f7055ac24a5c9b6e53d2510c9 --- /dev/null +++ b/mlp_out_all/layer_3_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_4_width_16k_l0_big/config.json b/mlp_out_all/layer_4_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8858f5cd765df1b843dfbeac2ecc6f7aea3913d5 --- /dev/null +++ b/mlp_out_all/layer_4_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 100, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_4_width_16k_l0_small/config.json b/mlp_out_all/layer_4_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b6cf3a705271000d4237d32223bb9547cf1945d0 --- /dev/null +++ b/mlp_out_all/layer_4_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_4_width_262k_l0_big/config.json b/mlp_out_all/layer_4_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3c837e095af0617ea03b8ea3e962c8000b7a4478 --- /dev/null +++ b/mlp_out_all/layer_4_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 100, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_4_width_262k_l0_small/config.json b/mlp_out_all/layer_4_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4db1a192352450ff43e486221a6d17d754273851 --- /dev/null +++ b/mlp_out_all/layer_4_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_5_width_16k_l0_big/config.json b/mlp_out_all/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..657ab9702d8a6efab69c62b341f1c6af6c8a35d5 --- /dev/null +++ b/mlp_out_all/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 110, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_5_width_16k_l0_small/config.json b/mlp_out_all/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..30b468474c5f93e9ba772fa4fbf3dab139ef0e28 --- /dev/null +++ b/mlp_out_all/layer_5_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_5_width_262k_l0_big/config.json b/mlp_out_all/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a5db984e287954684b0cf38a6add2b6a8e61c21b --- /dev/null +++ b/mlp_out_all/layer_5_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 110, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_5_width_262k_l0_small/config.json b/mlp_out_all/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c6d871f7619bcb665c395af556cc19c756025dfa --- /dev/null +++ b/mlp_out_all/layer_5_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_6_width_16k_l0_big/config.json b/mlp_out_all/layer_6_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6628380943582cdfafb237ede58aca81a528e4b3 --- /dev/null +++ b/mlp_out_all/layer_6_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_6_width_16k_l0_small/config.json b/mlp_out_all/layer_6_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4e617eadfa7562d031aa6f742c058ce38defe4f7 --- /dev/null +++ b/mlp_out_all/layer_6_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_6_width_262k_l0_big/config.json b/mlp_out_all/layer_6_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..04472343bb8888d24899e85bcdb7d977d10305f5 --- /dev/null +++ b/mlp_out_all/layer_6_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_6_width_262k_l0_small/config.json b/mlp_out_all/layer_6_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..69debce96460d3a7d556ef652fdade1505f0dd57 --- /dev/null +++ b/mlp_out_all/layer_6_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_7_width_16k_l0_big/config.json b/mlp_out_all/layer_7_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..de12228533662a6725fa285ab0e59ca4cceb5772 --- /dev/null +++ b/mlp_out_all/layer_7_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_7_width_16k_l0_small/config.json b/mlp_out_all/layer_7_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07b0878977592a5642aa5eaf69f4a2a32c32d975 --- /dev/null +++ b/mlp_out_all/layer_7_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_7_width_262k_l0_big/config.json b/mlp_out_all/layer_7_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9c9fabcd6f2a9bd180c77bdfbedf635a78243f89 --- /dev/null +++ b/mlp_out_all/layer_7_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_7_width_262k_l0_small/config.json b/mlp_out_all/layer_7_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f27b01aace758faec6a920f579917e79e3695aa0 --- /dev/null +++ b/mlp_out_all/layer_7_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_8_width_16k_l0_big/config.json b/mlp_out_all/layer_8_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c8bb5846079450ad8b9f5716fa6478dd45ede8f5 --- /dev/null +++ b/mlp_out_all/layer_8_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_8_width_16k_l0_small/config.json b/mlp_out_all/layer_8_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8b365f205d47f264bb5137eae09217d7b8b73831 --- /dev/null +++ b/mlp_out_all/layer_8_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_8_width_262k_l0_big/config.json b/mlp_out_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..82cf14d2a44f0bcf6e40f04908067d2eec89fa7b --- /dev/null +++ b/mlp_out_all/layer_8_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_8_width_262k_l0_small/config.json b/mlp_out_all/layer_8_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b79bee202e3b5510b6ed3b9f9a2ca2843e16c6ca --- /dev/null +++ b/mlp_out_all/layer_8_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_9_width_16k_l0_big/config.json b/mlp_out_all/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e620cbcd3a6e36760b4f8cf1eee7c7c8a0bba851 --- /dev/null +++ b/mlp_out_all/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_9_width_16k_l0_small/config.json b/mlp_out_all/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39288522b0ffab750eb98e79b80f57284e5c95a5 --- /dev/null +++ b/mlp_out_all/layer_9_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_9_width_262k_l0_big/config.json b/mlp_out_all/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1ac7c41c4a5adcd9d64d929464ed0d67468dd9e8 --- /dev/null +++ b/mlp_out_all/layer_9_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_9_width_262k_l0_small/config.json b/mlp_out_all/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7fce4e84f82961fd495e31fdaf1f75f8c2541383 --- /dev/null +++ b/mlp_out_all/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_10_width_16k_l0_big/config.json b/resid_post_all/layer_10_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..277ada1c96b795924bd1e0b8c56f0d18d597551d --- /dev/null +++ b/resid_post_all/layer_10_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.output", + "hf_hook_point_out": "model.layers.10.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_11_width_262k_l0_small/config.json b/resid_post_all/layer_11_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84a8d7e1836ad02189b1cde831b2a20b43c76706 --- /dev/null +++ b/resid_post_all/layer_11_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.output", + "hf_hook_point_out": "model.layers.11.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_17_width_16k_l0_big/config.json b/resid_post_all/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..177cdd23a914230c2ad9fe919b1daa9ba337f9aa --- /dev/null +++ b/resid_post_all/layer_17_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_2_width_262k_l0_big/config.json b/resid_post_all/layer_2_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c36927d065c7fe8e44c83fe8e63d3f30d2e0bd5c --- /dev/null +++ b/resid_post_all/layer_2_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.output", + "hf_hook_point_out": "model.layers.2.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_2_width_262k_l0_small/config.json b/resid_post_all/layer_2_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..33d22dd5e309b8dcc4e327ad88318da343952736 --- /dev/null +++ b/resid_post_all/layer_2_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.output", + "hf_hook_point_out": "model.layers.2.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_7_width_262k_l0_small/config.json b/resid_post_all/layer_7_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..99e50b6fd9e614ab9f2a18b44e194c047d480558 --- /dev/null +++ b/resid_post_all/layer_7_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.output", + "hf_hook_point_out": "model.layers.7.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_9_width_16k_l0_small/config.json b/resid_post_all/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0ce09979536a93a624d3a08e236e501ade40a241 --- /dev/null +++ b/resid_post_all/layer_9_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file