diff --git a/attn_out/layer_16_width_16k_l0_big/config.json b/attn_out/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2433fe205550c781ddb0e44bd19644b49022fe8a --- /dev/null +++ b/attn_out/layer_16_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_16k_l0_medium/config.json b/attn_out/layer_16_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d88b1ec8f35fff465478e60bf4d9576483333aa --- /dev/null +++ b/attn_out/layer_16_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_16k_l0_small/config.json b/attn_out/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..15ef4ef16ff3837570f9ad08cdd59369788dc950 --- /dev/null +++ b/attn_out/layer_16_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_262k_l0_big/config.json b/attn_out/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..de9c34c83473746feda59cc757361354a9f99cc6 --- /dev/null +++ b/attn_out/layer_16_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_262k_l0_medium/config.json b/attn_out/layer_16_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..52f38098938252c4efd120e28320a45c9ece2fb5 --- /dev/null +++ b/attn_out/layer_16_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_262k_l0_small/config.json b/attn_out/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..23880b3de71370357754e937ed7e09f8b30aff61 --- /dev/null +++ b/attn_out/layer_16_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_65k_l0_big/config.json b/attn_out/layer_16_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..335bad84b0d0804b5969d472e27689ce328cc0a8 --- /dev/null +++ b/attn_out/layer_16_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_65k_l0_medium/config.json b/attn_out/layer_16_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f65846703fa9d5d2c65e71ad66e5748178e1de2c --- /dev/null +++ b/attn_out/layer_16_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_16_width_65k_l0_small/config.json b/attn_out/layer_16_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..858b7db8915b9a08e4474e6c8f642be1fc575877 --- /dev/null +++ b/attn_out/layer_16_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.16.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_16k_l0_big/config.json b/attn_out/layer_31_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..16974e93396c936ffa697d7f66d565d1a4738e84 --- /dev/null +++ b/attn_out/layer_31_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_16k_l0_medium/config.json b/attn_out/layer_31_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b3000041e637803615ad74e4ce20846a60f1db36 --- /dev/null +++ b/attn_out/layer_31_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_16k_l0_small/config.json b/attn_out/layer_31_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2df7c90969e5c4c4716f909b8be6d88002df0476 --- /dev/null +++ b/attn_out/layer_31_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_262k_l0_big/config.json b/attn_out/layer_31_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f7ad8025bba4cc41eeaec37c7722de4c586ea5e8 --- /dev/null +++ b/attn_out/layer_31_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_262k_l0_medium/config.json b/attn_out/layer_31_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a78d01fa1ab94e9ddb0b67e167079374d8b03120 --- /dev/null +++ b/attn_out/layer_31_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_262k_l0_small/config.json b/attn_out/layer_31_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a0329e4cd8edcd07e2afcfc36b3b6e2b3a100be0 --- /dev/null +++ b/attn_out/layer_31_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_65k_l0_big/config.json b/attn_out/layer_31_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1adc54e77a40bebd73257d1442b1417cad0db311 --- /dev/null +++ b/attn_out/layer_31_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_65k_l0_medium/config.json b/attn_out/layer_31_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..14ce5adbc42c0e9d60168652e4c0ed072645bead --- /dev/null +++ b/attn_out/layer_31_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_31_width_65k_l0_small/config.json b/attn_out/layer_31_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..38a64eface93149810be62cdb759af0211f48d51 --- /dev/null +++ b/attn_out/layer_31_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.31.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_16k_l0_big/config.json b/attn_out/layer_40_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..36fe549455070b317fab59714317ae5bad87949f --- /dev/null +++ b/attn_out/layer_40_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_16k_l0_medium/config.json b/attn_out/layer_40_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..56474774d3e55eeac70f3a0c08408689f68f832c --- /dev/null +++ b/attn_out/layer_40_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_16k_l0_small/config.json b/attn_out/layer_40_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..230be4ba70a35b582a9d1032f47ee156baa0224a --- /dev/null +++ b/attn_out/layer_40_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_262k_l0_big/config.json b/attn_out/layer_40_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..82ac1f537326f8abe592a106e4591c9c8d85fa95 --- /dev/null +++ b/attn_out/layer_40_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_262k_l0_medium/config.json b/attn_out/layer_40_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..400c35de99e3e2610b1e589b4b4049598e75eb0a --- /dev/null +++ b/attn_out/layer_40_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_262k_l0_small/config.json b/attn_out/layer_40_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bf76cc08ead0fa6c1fae09c261b761c6428655a9 --- /dev/null +++ b/attn_out/layer_40_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_65k_l0_big/config.json b/attn_out/layer_40_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12da005975ea306bf78dc538ee0ed40d8edbce9f --- /dev/null +++ b/attn_out/layer_40_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_65k_l0_medium/config.json b/attn_out/layer_40_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..401cd2e7c756e9fa17d85c8047673a6db8b2cad3 --- /dev/null +++ b/attn_out/layer_40_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_40_width_65k_l0_small/config.json b/attn_out/layer_40_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ac94a09b61e607e4f788feb777efd6c8edddd145 --- /dev/null +++ b/attn_out/layer_40_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.40.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_16k_l0_big/config.json b/attn_out/layer_53_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ed83508465b5c09caeda5b9f91b87de45040015 --- /dev/null +++ b/attn_out/layer_53_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_16k_l0_medium/config.json b/attn_out/layer_53_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5e401c25c356375167d597104842ee899d318e7b --- /dev/null +++ b/attn_out/layer_53_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_16k_l0_small/config.json b/attn_out/layer_53_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ce7936eb79bc66b204134e0732f36070ebe83958 --- /dev/null +++ b/attn_out/layer_53_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_262k_l0_big/config.json b/attn_out/layer_53_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..15144114994b8bab7544fcaceb07e95458ad7264 --- /dev/null +++ b/attn_out/layer_53_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_262k_l0_medium/config.json b/attn_out/layer_53_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..26faf3c6986a9df4ab69c44992fc2a526f1f7b61 --- /dev/null +++ b/attn_out/layer_53_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_262k_l0_small/config.json b/attn_out/layer_53_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9dc3d8623e8f56781ecb21c3c7a1bfa3a83ef2a6 --- /dev/null +++ b/attn_out/layer_53_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_65k_l0_big/config.json b/attn_out/layer_53_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c63589c9a5e9404a2576a0209a39eb231d29655c --- /dev/null +++ b/attn_out/layer_53_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_65k_l0_medium/config.json b/attn_out/layer_53_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a2efaf353a1c312f80a426c5064dccbcf0f6902a --- /dev/null +++ b/attn_out/layer_53_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out/layer_53_width_65k_l0_small/config.json b/attn_out/layer_53_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..884ebb5ffa0b5779b10d5cc3e75f0bbe80d61b56 --- /dev/null +++ b/attn_out/layer_53_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.53.self_attn.o_proj.input", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_12_width_16k_l0_small/config.json b/attn_out_all/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fdceb053ebf95da328a86e01f5c36d759f42ad82 --- /dev/null +++ b/attn_out_all/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.12.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_17_width_16k_l0_big/config.json b/attn_out_all/layer_17_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..236c9d5d90f5c036aa073d993ac4ce65435025b1 --- /dev/null +++ b/attn_out_all/layer_17_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.17.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 109, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_23_width_262k_l0_big/config.json b/attn_out_all/layer_23_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9851a62f4fd59be432830ad85b5c0105f1172af2 --- /dev/null +++ b/attn_out_all/layer_23_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.23.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.23.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_24_width_16k_l0_big/config.json b/attn_out_all/layer_24_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4053093731086e2924d8a749721487e2e43e79d6 --- /dev/null +++ b/attn_out_all/layer_24_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.24.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.24.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_32_width_16k_l0_big/config.json b/attn_out_all/layer_32_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b4d9c13b62f3450cacbce2b0e2b7cda28e3932ab --- /dev/null +++ b/attn_out_all/layer_32_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.32.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_32_width_262k_l0_small/config.json b/attn_out_all/layer_32_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9f33c9b4d88a67787df501b11979b17c8d7b5b9e --- /dev/null +++ b/attn_out_all/layer_32_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.32.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.32.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_33_width_262k_l0_big/config.json b/attn_out_all/layer_33_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7367896e329e18c58b03c64b9947f4a6a9c68cc9 --- /dev/null +++ b/attn_out_all/layer_33_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.33.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_33_width_262k_l0_small/config.json b/attn_out_all/layer_33_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..31727fccd5c51990113e6c2ccec8e9189c4f81b4 --- /dev/null +++ b/attn_out_all/layer_33_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.33.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_37_width_262k_l0_small/config.json b/attn_out_all/layer_37_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d9600b3eeac6ba369fed99e67d65d5c49eec0868 --- /dev/null +++ b/attn_out_all/layer_37_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.37.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.37.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_38_width_262k_l0_small/config.json b/attn_out_all/layer_38_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e252930cdb5b91d88033eb0906d998e6dc20c2c8 --- /dev/null +++ b/attn_out_all/layer_38_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.38.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_41_width_262k_l0_small/config.json b/attn_out_all/layer_41_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..72b32c64c7fbe80d92d7e8ca8a2ce49e4a8e28ea --- /dev/null +++ b/attn_out_all/layer_41_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.41.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.41.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_42_width_16k_l0_small/config.json b/attn_out_all/layer_42_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e652869648edb10e5b324a808f2f8cddb8663c63 --- /dev/null +++ b/attn_out_all/layer_42_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.42.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.42.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_46_width_16k_l0_big/config.json b/attn_out_all/layer_46_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..42ee0c328d3e4dc990c2de3f085f1ac5fac8c543 --- /dev/null +++ b/attn_out_all/layer_46_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.46.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_46_width_262k_l0_small/config.json b/attn_out_all/layer_46_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..443ecb86335343b2cf452e005d4f251159fd59ad --- /dev/null +++ b/attn_out_all/layer_46_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.46.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.46.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_48_width_16k_l0_big/config.json b/attn_out_all/layer_48_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3faed14fc6c7c262da89ff3376b6251bbe9ceca3 --- /dev/null +++ b/attn_out_all/layer_48_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.48.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.48.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_51_width_262k_l0_small/config.json b/attn_out_all/layer_51_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3bde3bdd98904732ba599fc94651e6cebf1bb73a --- /dev/null +++ b/attn_out_all/layer_51_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.51.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.51.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_52_width_16k_l0_big/config.json b/attn_out_all/layer_52_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..24af38771fd5a5fff27b22d5cb59f59e25ef61ed --- /dev/null +++ b/attn_out_all/layer_52_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.52.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.52.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_56_width_16k_l0_big/config.json b/attn_out_all/layer_56_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..70f5cd687ca25d652e6fbe8da098768f133771e5 --- /dev/null +++ b/attn_out_all/layer_56_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.56.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.56.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_59_width_262k_l0_big/config.json b/attn_out_all/layer_59_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..563a964b6c777b3db3fa271f852d7d67b660d8c3 --- /dev/null +++ b/attn_out_all/layer_59_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.59.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.59.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_61_width_16k_l0_small/config.json b/attn_out_all/layer_61_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..86d04e1df0dadaa8dec5f91eb0981a64ae222176 --- /dev/null +++ b/attn_out_all/layer_61_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.61.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.61.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_6_width_262k_l0_big/config.json b/attn_out_all/layer_6_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a666a204fc0ba02778b6d41d7ce66e8854f80c1b --- /dev/null +++ b/attn_out_all/layer_6_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 77, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_6_width_262k_l0_small/config.json b/attn_out_all/layer_6_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ba61f1cc8aa34e603c3b3a1c73ac20ca5b58726 --- /dev/null +++ b/attn_out_all/layer_6_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.6.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 12, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_7_width_262k_l0_big/config.json b/attn_out_all/layer_7_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a04ace3c51f21342d113f3ac67e827311fad227a --- /dev/null +++ b/attn_out_all/layer_7_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.7.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_16k_l0_big/config.json b/mlp_out/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fb6b294e50a733f885f52da3aab1c54a5a684f80 --- /dev/null +++ b/mlp_out/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_16k_l0_medium/config.json b/mlp_out/layer_16_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..44a8b3ada4c11ca8d538ac8076d65513c06490d1 --- /dev/null +++ b/mlp_out/layer_16_width_16k_l0_medium/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_16k_l0_small/config.json b/mlp_out/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3df020ed20f748eff768d97db7754fc9fe4f23dd --- /dev/null +++ b/mlp_out/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_262k_l0_big/config.json b/mlp_out/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ffd00e9818293b65671eac43a9a2a7a247d62d8f --- /dev/null +++ b/mlp_out/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_262k_l0_medium/config.json b/mlp_out/layer_16_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..923dda4f2f4c8ad123329a7f76dd1d68026174c6 --- /dev/null +++ b/mlp_out/layer_16_width_262k_l0_medium/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_262k_l0_small/config.json b/mlp_out/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3cc7f3d45926f6600085effb69d6ad57f79f38a4 --- /dev/null +++ b/mlp_out/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-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_65k_l0_big/config.json b/mlp_out/layer_16_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4a3c1486694c3ccd6e608a7ea2f55282e95787a6 --- /dev/null +++ b/mlp_out/layer_16_width_65k_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": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 129, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_65k_l0_medium/config.json b/mlp_out/layer_16_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ffcebb05ad1b93f7739d4a8d7d1e0b344ca33c0c --- /dev/null +++ b/mlp_out/layer_16_width_65k_l0_medium/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": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 53, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_16_width_65k_l0_small/config.json b/mlp_out/layer_16_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..acece86257c93c4846b18af38487eedee6e5643e --- /dev/null +++ b/mlp_out/layer_16_width_65k_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": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 17, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_16k_l0_big/config.json b/mlp_out/layer_31_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84e0798c036dae4f4065f543bd05b8c25e8fe9e2 --- /dev/null +++ b/mlp_out/layer_31_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_16k_l0_medium/config.json b/mlp_out/layer_31_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e5256e17f2caeff255d4bf80bc54a86da4d09a1b --- /dev/null +++ b/mlp_out/layer_31_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_16k_l0_small/config.json b/mlp_out/layer_31_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c27f3f282e86e5e95f19f77ac69dd25fe31c661 --- /dev/null +++ b/mlp_out/layer_31_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_262k_l0_big/config.json b/mlp_out/layer_31_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7d2ae7dace1a731aa9eee124b61c3e8a37a17035 --- /dev/null +++ b/mlp_out/layer_31_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_262k_l0_medium/config.json b/mlp_out/layer_31_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..149ab54729834f035c1e594cca998ed0c41df94c --- /dev/null +++ b/mlp_out/layer_31_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_262k_l0_small/config.json b/mlp_out/layer_31_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e34cc9d6b29900f6b193a1c93c47b130c0d59e7 --- /dev/null +++ b/mlp_out/layer_31_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_65k_l0_big/config.json b/mlp_out/layer_31_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f5580888e5e8d5fc712b1c67cf36e7b607bfbfc8 --- /dev/null +++ b/mlp_out/layer_31_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_65k_l0_medium/config.json b/mlp_out/layer_31_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a34fc8f6211e429f06814af1e7fbdbfb3b1375f3 --- /dev/null +++ b/mlp_out/layer_31_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_31_width_65k_l0_small/config.json b/mlp_out/layer_31_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a68dc431e445e6d071a5e18ad848d38a7760c27 --- /dev/null +++ b/mlp_out/layer_31_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_16k_l0_big/config.json b/mlp_out/layer_40_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ee03feb504e790c5e80e54979b626048f119296e --- /dev/null +++ b/mlp_out/layer_40_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_16k_l0_medium/config.json b/mlp_out/layer_40_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dd5531e4e30cac49b4697965ece3eb579466329b --- /dev/null +++ b/mlp_out/layer_40_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_16k_l0_small/config.json b/mlp_out/layer_40_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bfda1c98574ac7aa64ee5247909a2092a04df4b7 --- /dev/null +++ b/mlp_out/layer_40_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_262k_l0_big/config.json b/mlp_out/layer_40_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f893e19098023c93451e14087245fbaabac84c4 --- /dev/null +++ b/mlp_out/layer_40_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_262k_l0_medium/config.json b/mlp_out/layer_40_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..df9bea1f667bf872b82f1aaf9a833936eeedc784 --- /dev/null +++ b/mlp_out/layer_40_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_262k_l0_small/config.json b/mlp_out/layer_40_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e554813bd103fb925e9c39cad9cd72cb04cf873 --- /dev/null +++ b/mlp_out/layer_40_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_65k_l0_big/config.json b/mlp_out/layer_40_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a48b2f56c3baad0b09453b14af27c4fbe71c244b --- /dev/null +++ b/mlp_out/layer_40_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_65k_l0_medium/config.json b/mlp_out/layer_40_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..be21a546a57975002d751f9daaabf53bb63daa8a --- /dev/null +++ b/mlp_out/layer_40_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_40_width_65k_l0_small/config.json b/mlp_out/layer_40_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ab78088921af301c8af857ced9043976ddbbb1a --- /dev/null +++ b/mlp_out/layer_40_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_16k_l0_big/config.json b/mlp_out/layer_53_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..149d5af3c390918a6d89a4c9be5a3bc0d9a86159 --- /dev/null +++ b/mlp_out/layer_53_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_16k_l0_medium/config.json b/mlp_out/layer_53_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ce097ccff6d4b99dcdbdde7636eaa78bc2500c53 --- /dev/null +++ b/mlp_out/layer_53_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_16k_l0_small/config.json b/mlp_out/layer_53_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..065a73e3c064312237c7ee86814bfca8ea76699e --- /dev/null +++ b/mlp_out/layer_53_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_262k_l0_big/config.json b/mlp_out/layer_53_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f2e3cfd454e8c1fe1ca1cfb4fcd9f02dc81a783c --- /dev/null +++ b/mlp_out/layer_53_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_262k_l0_medium/config.json b/mlp_out/layer_53_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..38405c62473823e696492330a17c9bbed5da761d --- /dev/null +++ b/mlp_out/layer_53_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_262k_l0_small/config.json b/mlp_out/layer_53_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5ab36de4ee5161568b07e241306ec43e59e63d7 --- /dev/null +++ b/mlp_out/layer_53_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_65k_l0_big/config.json b/mlp_out/layer_53_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5dfb721c4baac0578a493cb6961c50d01f1da2a6 --- /dev/null +++ b/mlp_out/layer_53_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_65k_l0_medium/config.json b/mlp_out/layer_53_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3fdbc24028b14ba02af34099d578a2f467745aeb --- /dev/null +++ b/mlp_out/layer_53_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out/layer_53_width_65k_l0_small/config.json b/mlp_out/layer_53_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..eb6ff03811648f2620d4ad15ff8c2fb88c59ee74 --- /dev/null +++ b/mlp_out/layer_53_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.53.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.53.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "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..eff16ea46446a24331a74663565e61ed250f5fef --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 60, + "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..09368b6ffd7f4ceb7e51176695c4fe3091b3b867 --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 100, + "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..88da426a2c742edb422ac0dfac34303dbb914ac9 --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 103, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_18_width_16k_l0_big/config.json b/mlp_out_all/layer_18_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7402ddf6933e656a0a915f5df72aaf9a63558fa7 --- /dev/null +++ b/mlp_out_all/layer_18_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.18.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.18.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 112, + "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..257b0bb606f76a1195d71f83c102c6beed518f2a --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_20_width_16k_l0_small/config.json b/mlp_out_all/layer_20_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec2e3bad840dc005da890255ae7f9b9d4bfbb7fb --- /dev/null +++ b/mlp_out_all/layer_20_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.20.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.20.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 19, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_31_width_16k_l0_big/config.json b/mlp_out_all/layer_31_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ee28ebf49791ab4c88440bcfb9532182dcc92cf8 --- /dev/null +++ b/mlp_out_all/layer_31_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.31.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.31.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_33_width_16k_l0_big/config.json b/mlp_out_all/layer_33_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f843d2383e79587757ea1b17061e3a32a3138c2 --- /dev/null +++ b/mlp_out_all/layer_33_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.33.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.33.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_36_width_262k_l0_small/config.json b/mlp_out_all/layer_36_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..604a90238d249ce8e07010bba03e00fa90707b81 --- /dev/null +++ b/mlp_out_all/layer_36_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.36.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.36.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_38_width_16k_l0_big/config.json b/mlp_out_all/layer_38_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1a9127e5cff1e1f89c9a03b60677dd44d9324dff --- /dev/null +++ b/mlp_out_all/layer_38_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.38.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.38.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_40_width_262k_l0_small/config.json b/mlp_out_all/layer_40_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e554813bd103fb925e9c39cad9cd72cb04cf873 --- /dev/null +++ b/mlp_out_all/layer_40_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.40.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.40.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_43_width_262k_l0_big/config.json b/mlp_out_all/layer_43_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6edb9b396f9705b6c5a9a8bc1fa304734deeb499 --- /dev/null +++ b/mlp_out_all/layer_43_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.43.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.43.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_45_width_262k_l0_big/config.json b/mlp_out_all/layer_45_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7d67f7801f72a2d45a580938338208337215d247 --- /dev/null +++ b/mlp_out_all/layer_45_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_45_width_262k_l0_small/config.json b/mlp_out_all/layer_45_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1e563c010baecf8ebfa14ee17d6327a6ff63f000 --- /dev/null +++ b/mlp_out_all/layer_45_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.45.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.45.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_47_width_16k_l0_big/config.json b/mlp_out_all/layer_47_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6f751ecbef5fd2d48aef3c5ad28b596bbf3d4aa3 --- /dev/null +++ b/mlp_out_all/layer_47_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_47_width_262k_l0_small/config.json b/mlp_out_all/layer_47_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ccea8223556600f04f892c701ce7e42a52d72976 --- /dev/null +++ b/mlp_out_all/layer_47_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.47.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.47.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_49_width_262k_l0_big/config.json b/mlp_out_all/layer_49_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..506ca2b238beba630e744e42107dbfe45684358c --- /dev/null +++ b/mlp_out_all/layer_49_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.49.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.49.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_50_width_262k_l0_big/config.json b/mlp_out_all/layer_50_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3594879f5ca68eaed5293785e68d8c48e6d7effe --- /dev/null +++ b/mlp_out_all/layer_50_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.50.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.50.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_54_width_16k_l0_small/config.json b/mlp_out_all/layer_54_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dd3cd82050529e163808a4c5604f8b7de9eb05ea --- /dev/null +++ b/mlp_out_all/layer_54_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.54.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.54.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_55_width_16k_l0_big/config.json b/mlp_out_all/layer_55_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db16c8b16114483d3bcfec2401911f7e685e3c95 --- /dev/null +++ b/mlp_out_all/layer_55_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.55.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.55.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_58_width_262k_l0_big/config.json b/mlp_out_all/layer_58_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..54b4f55e5ba3599951233728b046e4237bacd496 --- /dev/null +++ b/mlp_out_all/layer_58_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.58.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.58.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 120, + "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..1ca5bce9c4147fb62ed704e964d5e96a0a8481dd --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 74, + "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..da867b878ac0d25f6f8e8022ef20949cd682d858 --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 74, + "affine_connection": false +} \ No newline at end of file diff --git a/mlp_out_all/layer_60_width_16k_l0_small/config.json b/mlp_out_all/layer_60_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0c0a2eb3bbf52ee91e2bd0a38163398756a2995 --- /dev/null +++ b/mlp_out_all/layer_60_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.60.post_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.60.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-27b-it", + "architecture": "jump_relu", + "l0": 20, + "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..d627db9b366633117ba8596c3143796456811803 --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 12, + "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..d4df066b8a5923629b7c6ff3735d2128a2a20213 --- /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-v3-27b-it", + "architecture": "jump_relu", + "l0": 86, + "affine_connection": false +} \ No newline at end of file