diff --git a/attn_out_all/layer_10_width_262k_l0_small/config.json b/attn_out_all/layer_10_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b5dfcc498e9a073b53581e174cd32f72214d5d24 --- /dev/null +++ b/attn_out_all/layer_10_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.10.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_11_width_262k_l0_big/config.json b/attn_out_all/layer_11_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7bce2b80f4852d1937280ba305080b44c7d1897a --- /dev/null +++ b/attn_out_all/layer_11_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.11.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_12_width_16k_l0_big/config.json b/attn_out_all/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aca5daba0aa0c9b10843f953ea6dda9f7d8e2646 --- /dev/null +++ b/attn_out_all/layer_12_width_16k_l0_big/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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "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..29b5ac659ab546465baaa5dc1977741db90c4219 --- /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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_12_width_262k_l0_big/config.json b/attn_out_all/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2e1fc5e1b6928c0259588d09308b27a73717c0b5 --- /dev/null +++ b/attn_out_all/layer_12_width_262k_l0_big/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": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_12_width_262k_l0_small/config.json b/attn_out_all/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3da5a1b1e67e845033f8054af063c1f0aade6a3d --- /dev/null +++ b/attn_out_all/layer_12_width_262k_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": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_13_width_262k_l0_big/config.json b/attn_out_all/layer_13_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..afaa6e3a12418945c257001d93d4a69cb0f5df22 --- /dev/null +++ b/attn_out_all/layer_13_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.13.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_13_width_262k_l0_small/config.json b/attn_out_all/layer_13_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0b020cd7fe05a4eb7232a61f124cdb0ecc58cb83 --- /dev/null +++ b/attn_out_all/layer_13_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.13.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_14_width_16k_l0_small/config.json b/attn_out_all/layer_14_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b3ea3ab29c169fdeb38197f33bfcf1ce31acd560 --- /dev/null +++ b/attn_out_all/layer_14_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.14.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_14_width_262k_l0_big/config.json b/attn_out_all/layer_14_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..283b16fe1ebeb1e480f6e90cee116fc9c3be1315 --- /dev/null +++ b/attn_out_all/layer_14_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.14.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_15_width_16k_l0_small/config.json b/attn_out_all/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8af587587fbb3f3aff89e019e7c99354605214f0 --- /dev/null +++ b/attn_out_all/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.15.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_15_width_262k_l0_small/config.json b/attn_out_all/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ab9c7f82922964f336399667721e6e876c7b624 --- /dev/null +++ b/attn_out_all/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.15.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_16_width_16k_l0_big/config.json b/attn_out_all/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0d849adf7f6d0523400e9e75a2206fe37127c13b --- /dev/null +++ b/attn_out_all/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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_16_width_16k_l0_small/config.json b/attn_out_all/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ecd49d43cef32d352a1f1d4da0ff2eeedf149039 --- /dev/null +++ b/attn_out_all/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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_16_width_262k_l0_big/config.json b/attn_out_all/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c6aa890332e4784c7d15cdb5341b80f9be6d384d --- /dev/null +++ b/attn_out_all/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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_1_width_16k_l0_big/config.json b/attn_out_all/layer_1_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4327e3478d0e536b72e1d2804755a06455b01fed --- /dev/null +++ b/attn_out_all/layer_1_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_1_width_16k_l0_small/config.json b/attn_out_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..92c892b8fee9b04681b2a7030b16334e4b3506cd --- /dev/null +++ b/attn_out_all/layer_1_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_1_width_262k_l0_big/config.json b/attn_out_all/layer_1_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..578acbf6e999cf6a0866c66932d4730e9415832e --- /dev/null +++ b/attn_out_all/layer_1_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_1_width_262k_l0_small/config.json b/attn_out_all/layer_1_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d4015821e11c3bdb61f2eac53710e5682d8e89c --- /dev/null +++ b/attn_out_all/layer_1_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.1.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_3_width_262k_l0_big/config.json b/attn_out_all/layer_3_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a14551633380e97df56ad496aede61f888883cf0 --- /dev/null +++ b/attn_out_all/layer_3_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.3.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_4_width_262k_l0_small/config.json b/attn_out_all/layer_4_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9dfb2d5a8fb0d6c93ccaa1dd347457ec647e08a7 --- /dev/null +++ b/attn_out_all/layer_4_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.4.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_5_width_262k_l0_small/config.json b/attn_out_all/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..13ce70a30ef2f5d4e4bb2dc4e91b034350ff9723 --- /dev/null +++ b/attn_out_all/layer_5_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.5.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_6_width_16k_l0_big/config.json b/attn_out_all/layer_6_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..be008c3bbb6763752ce5b858b3ce3ed5856f6745 --- /dev/null +++ b/attn_out_all/layer_6_width_16k_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": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "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..0b3a8e9a214642b9b8f51178b5f3679546426d42 --- /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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "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..dfe5f0eb86e6b1edd12422ae85a827073a5759fa --- /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-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_8_width_16k_l0_small/config.json b/attn_out_all/layer_8_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e552045de545f5c3b8dd77824b211c704e9fabe2 --- /dev/null +++ b/attn_out_all/layer_8_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.8.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_9_width_16k_l0_big/config.json b/attn_out_all/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..229756451ab14be5160c194e8d96580328cb75d1 --- /dev/null +++ b/attn_out_all/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/attn_out_all/layer_9_width_262k_l0_small/config.json b/attn_out_all/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..96caefc3b99925a09474b3f5cc6830a0e092e290 --- /dev/null +++ b/attn_out_all/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.self_attn.o_proj.input", + "hf_hook_point_out": "model.layers.9.self_attn.o_proj.input", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_262k_l0_big/config.json b/clt/width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d8f837ab9910be37f64163fb4e9c2bee5cdf9592 --- /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-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_262k_l0_medium/config.json b/clt/width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..14f09a47bafb91053dc1f4416c2c4973e5d2ddd8 --- /dev/null +++ b/clt/width_262k_l0_medium/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-pt", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_262k_l0_medium_affine/config.json b/clt/width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2990e11bcd0483b1947268297618133cbb8d9fb0 --- /dev/null +++ b/clt/width_262k_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": 262080, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": true +} \ No newline at end of file diff --git a/clt/width_524k_l0_medium/config.json b/clt/width_524k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6e8301abe54cb3e7bbc4e4be67b9b4b2ad86448 --- /dev/null +++ b/clt/width_524k_l0_medium/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-pt", + "architecture": "jump_relu", + "l0": 50, + "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..2d961d831b7dc562e38015d9478652e84799dfe9 --- /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-pt", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": true +} \ 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..5b39a4d16f0839fa34e6339d49bc67bc7e0045e0 --- /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-pt", + "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..89d6cba8572ca3d2e17d14ad6b1a2176d93c7e4e --- /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-pt", + "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..18d2faca08891efca7010f01c6620ecda47b2ab0 --- /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-pt", + "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..a4520b7657456db85278ed28bb611aaa180b432e --- /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-pt", + "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..ddfc741bb9af7aa6bef10ae61230f6146ef4b72e --- /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-pt", + "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..3d3b76ed2bf356072a5ca44d307fc94afb087702 --- /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-pt", + "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..f75a852c0e207847178327c3744e1fc5538369bf --- /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-pt", + "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_big/config.json b/mlp_out_all/layer_11_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ea92fa64db24c264806d312a21787b985d5c085 --- /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-pt", + "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..3cae67c396451bafb74d2e68c4f27df91447f4e0 --- /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-pt", + "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_small/config.json b/mlp_out_all/layer_11_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..36ac9f731b6119fbe8b17383c39666da3ef44324 --- /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-pt", + "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..e30c90aa2a45131be6cf2602265825aa5dd7ba54 --- /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-pt", + "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..43cec8ab7182619abc7e126b8ed53ea3b7cce855 --- /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-pt", + "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..0dc7c48fcee9f871cafb58ee31ecef4f9968951d --- /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-pt", + "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..2f386af075e39d4ee44a4cb16fa5a1ca0f974028 --- /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-pt", + "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..3225092e67831e863037f9a3493b47e725c730c2 --- /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-pt", + "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_big/config.json b/mlp_out_all/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e05ebb7eedc5f05b74f205cc57eec9a4495d4c07 --- /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-pt", + "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..ae6c353d1fc0bf3f6ee9d61dc3ec25e6d6969b84 --- /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-pt", + "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..f82d943a62eade9f904abfef3981971fe774de6f --- /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-pt", + "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..54b429fd6e3734584d462b921e02483b7b480988 --- /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-pt", + "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..fe8da18bbc2f6dad3abd1be835c5431cd188c28c --- /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-pt", + "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..c58764f4c8a78c4305bb88f8a094be77771c4bb0 --- /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-pt", + "architecture": "jump_relu", + "l0": 20, + "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..6114a6632864223b561e9bf5fc5f8356ed37788a --- /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-pt", + "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..ed4ff49c1924675f40029f9e6a0ec8e72cd657c7 --- /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-pt", + "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..8d81d9653bf67f232512584f3f94b59aa6fded49 --- /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-pt", + "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..40a1faf44bb4dddfa9d8998db81d938b586a727a --- /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-pt", + "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..88f295d1cae2758222f730858d76dd749857744f --- /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-pt", + "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..aed56bf8572aac6d3ec479389142211fce66936c --- /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-pt", + "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_small/config.json b/mlp_out_all/layer_3_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9f214248bab87c466f1108ae49318d99c580f8dc --- /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-pt", + "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..c5a5db4ccb7fd2dd2fc4575607576bb660a4eda9 --- /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-pt", + "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_big/config.json b/mlp_out_all/layer_4_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..956474b54126b2bd2e25eceb9799daaf01122776 --- /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-pt", + "architecture": "jump_relu", + "l0": 100, + "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..b01f07ee2638a1fefffbfd15f8a0ebb50042ed0e --- /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-pt", + "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..10784f014dd7bdd20db058ef0159a0bb74d016c1 --- /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-pt", + "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..382578d42b34b1afe3f7dd6fb9dbf2df032eeaa6 --- /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-pt", + "architecture": "jump_relu", + "l0": 110, + "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..9414c407e31cbd562577a503ab4557f533795066 --- /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-pt", + "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..1e1e6c31690c6ff2c94718ba164beb8e2c5c383f --- /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-pt", + "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..f3e9b45a7b109d135674e79174715e7b06aafe68 --- /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-pt", + "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..ee82f7f16da75b3c4d7e33aca2342a6b52984008 --- /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-pt", + "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_big/config.json b/mlp_out_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..18c9723c16b0cd7254e103947a9dc67c8e6ff3aa --- /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-pt", + "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..cd3b7db5c02977c5c6c0d825a2b1a85aee21c582 --- /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-pt", + "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..6b7233c46a9aa30cc672c61e59cd2956067107c2 --- /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-pt", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_16k_l0_big/config.json b/resid_post/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f826c237d10c98f68cb0794bad72f50ecbc6908c --- /dev/null +++ b/resid_post/layer_12_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_16k_l0_medium/config.json b/resid_post/layer_12_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84a8a70ce0e68403bd5eaf0a08ee9218460cac4c --- /dev/null +++ b/resid_post/layer_12_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_16k_l0_small/config.json b/resid_post/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bdbc41118ca80baa0635a1ffc26b3f343d2f1822 --- /dev/null +++ b/resid_post/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_1m_l0_big/config.json b/resid_post/layer_12_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bb9fd866d21b618a413e727da7d70653425f4b32 --- /dev/null +++ b/resid_post/layer_12_width_1m_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_1m_l0_medium/config.json b/resid_post/layer_12_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d214ae167d372115529633e229b266b0d568f029 --- /dev/null +++ b/resid_post/layer_12_width_1m_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_1m_l0_small/config.json b/resid_post/layer_12_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5f92826a843c66b967a062d868db1ab413ca783 --- /dev/null +++ b/resid_post/layer_12_width_1m_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_262k_l0_big/config.json b/resid_post/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2feee89b4d313aed046b55aab754d91ee8834210 --- /dev/null +++ b/resid_post/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_262k_l0_medium/config.json b/resid_post/layer_12_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..394af5d04044f1e6eab5caf80efebe5becf64376 --- /dev/null +++ b/resid_post/layer_12_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_12_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..394af5d04044f1e6eab5caf80efebe5becf64376 --- /dev/null +++ b/resid_post/layer_12_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_262k_l0_small/config.json b/resid_post/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e09713b37342c0c91afcb79f1289ad668278ae3f --- /dev/null +++ b/resid_post/layer_12_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_65k_l0_big/config.json b/resid_post/layer_12_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7973f9a3da19093bddc574bf1802d501419dc96b --- /dev/null +++ b/resid_post/layer_12_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_65k_l0_medium/config.json b/resid_post/layer_12_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1ff9e7286bee08eaa47a19011b4d19e4a39be5a5 --- /dev/null +++ b/resid_post/layer_12_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_12_width_65k_l0_small/config.json b/resid_post/layer_12_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a1af9ec0cc85e212b41a5d2acb6eb2e01de9b42e --- /dev/null +++ b/resid_post/layer_12_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.output", + "hf_hook_point_out": "model.layers.12.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_16k_l0_big/config.json b/resid_post/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6784bb2b3abd550103105b76a246ad9c309ab04c --- /dev/null +++ b/resid_post/layer_15_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_16k_l0_medium/config.json b/resid_post/layer_15_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..20ddc9179302742c112a98c32893b4b2d5abf671 --- /dev/null +++ b/resid_post/layer_15_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_16k_l0_small/config.json b/resid_post/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..697adb2126504cce797abd247a81e2e3434b8938 --- /dev/null +++ b/resid_post/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_1m_l0_big/config.json b/resid_post/layer_15_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0f9543d77a01237232f58f639243ec274d37e99a --- /dev/null +++ b/resid_post/layer_15_width_1m_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_1m_l0_medium/config.json b/resid_post/layer_15_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dcf2446de1bb9c1cfb7a4a1d23ad50806b00dd1f --- /dev/null +++ b/resid_post/layer_15_width_1m_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_1m_l0_small/config.json b/resid_post/layer_15_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b91407ff14d27b6aebc6d830c421de13c457dbf4 --- /dev/null +++ b/resid_post/layer_15_width_1m_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_262k_l0_big/config.json b/resid_post/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5006361837b2bef415de39ae72ad07a9987323a1 --- /dev/null +++ b/resid_post/layer_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_262k_l0_medium/config.json b/resid_post/layer_15_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e9c072836e3977f181b5013c87f40340355673c4 --- /dev/null +++ b/resid_post/layer_15_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_15_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e9c072836e3977f181b5013c87f40340355673c4 --- /dev/null +++ b/resid_post/layer_15_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_262k_l0_small/config.json b/resid_post/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c4a856a2116f3537365d984757e24a43e04f9a3 --- /dev/null +++ b/resid_post/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_65k_l0_big/config.json b/resid_post/layer_15_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4b4190d708841258f489facae074aada57e813c4 --- /dev/null +++ b/resid_post/layer_15_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_65k_l0_medium/config.json b/resid_post/layer_15_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c5850aab2af4c31ddb46d0dc48e4f6c37fd85202 --- /dev/null +++ b/resid_post/layer_15_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_15_width_65k_l0_small/config.json b/resid_post/layer_15_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5a42dea83a98348cf38da7992d88959f12dd9fe4 --- /dev/null +++ b/resid_post/layer_15_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.output", + "hf_hook_point_out": "model.layers.15.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_16k_l0_big/config.json b/resid_post/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0cae03f0dfdc8aae99f26430efa6e67a8021dd94 --- /dev/null +++ b/resid_post/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_16k_l0_medium/config.json b/resid_post/layer_5_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1dffbfd1225b4b8692d0215aae5ee21e848de185 --- /dev/null +++ b/resid_post/layer_5_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_16k_l0_small/config.json b/resid_post/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2dcb168cd43bc556091d381955ab6ffe1679e55f --- /dev/null +++ b/resid_post/layer_5_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 16384, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_1m_l0_big/config.json b/resid_post/layer_5_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..361289ab97b137fe0d3464d09219f43541b778fd --- /dev/null +++ b/resid_post/layer_5_width_1m_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_1m_l0_medium/config.json b/resid_post/layer_5_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..263c59adef870814494aa3a69fd38e1f53dd32d1 --- /dev/null +++ b/resid_post/layer_5_width_1m_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_1m_l0_small/config.json b/resid_post/layer_5_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0365e33dc70fb4a4b98762ca7362dee8ba374cce --- /dev/null +++ b/resid_post/layer_5_width_1m_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_262k_l0_big/config.json b/resid_post/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6b278a535fb9f367d255452c09f00df0669b5ee --- /dev/null +++ b/resid_post/layer_5_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_262k_l0_medium/config.json b/resid_post/layer_5_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d6091f0780e33b7fe149e88767e21b78db044b65 --- /dev/null +++ b/resid_post/layer_5_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_5_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d6091f0780e33b7fe149e88767e21b78db044b65 --- /dev/null +++ b/resid_post/layer_5_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_262k_l0_small/config.json b/resid_post/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..849efababe90effefabb5f6bdc8618ddc841ac81 --- /dev/null +++ b/resid_post/layer_5_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_65k_l0_big/config.json b/resid_post/layer_5_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..021ca208d942351de0743802539503a92eb86165 --- /dev/null +++ b/resid_post/layer_5_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_65k_l0_medium/config.json b/resid_post/layer_5_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..75a7c327505194ea035f1b23f0dafb3385be7bf7 --- /dev/null +++ b/resid_post/layer_5_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_5_width_65k_l0_small/config.json b/resid_post/layer_5_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bafe2fd212007b2e2e40c9669a186fe2a6bf2943 --- /dev/null +++ b/resid_post/layer_5_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.output", + "hf_hook_point_out": "model.layers.5.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_16k_l0_big/config.json b/resid_post/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..510f955d6bc6a71d115959026870bd9c378ceec9 --- /dev/null +++ b/resid_post/layer_9_width_16k_l0_big/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-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_16k_l0_medium/config.json b/resid_post/layer_9_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bfc4c337a6efa57f50fcea1ca9b538123968e693 --- /dev/null +++ b/resid_post/layer_9_width_16k_l0_medium/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-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_16k_l0_small/config.json b/resid_post/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4efe4128987bc9fc59773bc3055605c306e3dc1 --- /dev/null +++ b/resid_post/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-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_1m_l0_big/config.json b/resid_post/layer_9_width_1m_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1908e16649cf202d94531d45eaa22b38f43f0c21 --- /dev/null +++ b/resid_post/layer_9_width_1m_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_1m_l0_medium/config.json b/resid_post/layer_9_width_1m_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fa9c14d4535e9ad1686a31a363125a4de7edb3bb --- /dev/null +++ b/resid_post/layer_9_width_1m_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_1m_l0_small/config.json b/resid_post/layer_9_width_1m_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cbafe20704dbddeeeede6c3fb717c5c62b2c0136 --- /dev/null +++ b/resid_post/layer_9_width_1m_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": 1048576, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_262k_l0_big/config.json b/resid_post/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..28d0c5bac5e4e91df0e72d4a40e3679c13e35590 --- /dev/null +++ b/resid_post/layer_9_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_262k_l0_medium/config.json b/resid_post/layer_9_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..54205d2df209dbdda8a755e8e3e29970d8f62b5d --- /dev/null +++ b/resid_post/layer_9_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_262k_l0_medium_seed_1/config.json b/resid_post/layer_9_width_262k_l0_medium_seed_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..54205d2df209dbdda8a755e8e3e29970d8f62b5d --- /dev/null +++ b/resid_post/layer_9_width_262k_l0_medium_seed_1/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_262k_l0_small/config.json b/resid_post/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..faf2dc7cba6219565c0a30d5bd203411b2117486 --- /dev/null +++ b/resid_post/layer_9_width_262k_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": 262144, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_65k_l0_big/config.json b/resid_post/layer_9_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..66f89cda61cd32507c7cf79320bf00d202b443d1 --- /dev/null +++ b/resid_post/layer_9_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_65k_l0_medium/config.json b/resid_post/layer_9_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..909375d6bcbcd58529d78a74fa78b2c708342e26 --- /dev/null +++ b/resid_post/layer_9_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.output", + "hf_hook_point_out": "model.layers.9.output", + "width": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post/layer_9_width_65k_l0_small/config.json b/resid_post/layer_9_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3a96cc0fb63e55551b776e11357af073cc462e20 --- /dev/null +++ b/resid_post/layer_9_width_65k_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": 65536, + "model_name": "gemma-v3-270m-pt", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file