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..d1274cdbb737acd4ace70f8b2b68af73d3ec172c --- /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-it", + "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..a9a8200012daae69043a8eab5a48187e49b4e0be --- /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-it", + "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..59ef8231bffac97e6fb985ae5e8383d0e24335e8 --- /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-it", + "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..451276b588a740b897000a6570654372204d4fe3 --- /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-it", + "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..58e919fe7ff5f9c82998eb5caa2e1026d745da29 --- /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-it", + "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..741fb3dc840a023a549d7ba4556198f2d475fa5d --- /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-it", + "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..f225933e1bde6c9e7d968c5524992c9dffd95945 --- /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-it", + "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..31a3e469eacd4011bffaa851be004b80e638e173 --- /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-it", + "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..5ef435eafb914c722ff5e0113749a181dd2ef549 --- /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-it", + "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..1852942a325b4a757bad118595b61f5bedb8035a --- /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-it", + "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..e0690184806bca6fe8d8b4596c4d85e29c3b5172 --- /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-it", + "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..b6455588e9365da4c520afd98452d761e870d6de --- /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-it", + "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..40978811414c1bf8ac3b4a0dc4d6cb3cbdfd583d --- /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-it", + "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..ae3617d8a2b19abaf29a3af17e9b49a263d3396d --- /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-it", + "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..d607ebe4fb4c3416a2001f3155e3495c524faf32 --- /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-it", + "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..c2030dda03f4d9f6946904dbc9fb3394524b0954 --- /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-it", + "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..bfa9e89dfb7c58fd81f8de8fdc76aa3f2830d44b --- /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-it", + "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..a1d9cfcd2e6bdd778edfb2c1b632bd188a0119f5 --- /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-it", + "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..09129d98065f5c99dfcd801e2ddab48197ab5822 --- /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-it", + "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..f3be3a81a1717182d65d9ed2ed6c07ba5152d3a0 --- /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-it", + "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..fdc6610cd4be27ce3b6c06f99b56e39e53e2b2b7 --- /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-it", + "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..9163edbfbf5100058efdde4052df2a9efb460c68 --- /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-it", + "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..30f9ed5789d52af90ce459afe0b357d2916b74eb --- /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-it", + "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..cb8ba364caf862a161e5d73efb374bfbb2d101ac --- /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-it", + "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..2a593ff9036a33c9813d724ef7aec6ee360be28b --- /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-it", + "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..2a64da22f397bde265a918035026de30f4b72a03 --- /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-it", + "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..786553baa4846f264addadbf1e2a92839b762a96 --- /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-it", + "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..d95c406aef894f2f53481ddfcb0215cdd65214ea --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_262k_l0_big_affine/config.json b/clt/width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..245d8bccd2d7f458e4b9de967574f479e8c79bdf --- /dev/null +++ b/clt/width_262k_l0_big_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-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/clt/width_262k_l0_big_affine/params_layer_0.safetensors b/clt/width_262k_l0_big_affine/params_layer_0.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/clt/width_262k_l0_medium/config.json b/clt/width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a505d728e791c07a7398d9069f198c1e5a321c6d --- /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-it", + "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..ebef3c7344b3c85e6fc4d866591b5f01f8e9ab0d --- /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-it", + "architecture": "jump_relu", + "l0": 50, + "affine_connection": true +} \ No newline at end of file diff --git a/clt/width_262k_l0_medium_affine/params_layer_0.safetensors b/clt/width_262k_l0_medium_affine/params_layer_0.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/clt/width_524k_l0_big/config.json b/clt/width_524k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0f70342a47863834ff0a300ec55a5b1442b33c3e --- /dev/null +++ b/clt/width_524k_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": 524160, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/clt/width_524k_l0_big_affine/config.json b/clt/width_524k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c26ea295f9ae1260ee9de94c689d663e1feed190 --- /dev/null +++ b/clt/width_524k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.{all}.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.{all}.post_feedforward_layernorm.output", + "width": 524160, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "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..1e61cf4a1ef24ec1138f25f60ac9d7568bdfc0ca --- /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-it", + "architecture": "jump_relu", + "l0": 50, + "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..3683132d85c4dcc3b3edcc011cba9911c893ff25 --- /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-it", + "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..39ce4bba4f4323b0f7dafea47d5dae9daee908a6 --- /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-it", + "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..44fb5bb602ef875ca5c62c99117148ed6f4d54ea --- /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-it", + "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..5bdfb30b869df64d59e2dfc70d61f5fe71ba3632 --- /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-it", + "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..91c41dd0350bd96c399c93e60733d585d36c1d11 --- /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-it", + "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..77a82f0821b7c2ce7c45a02c9562c86e4730e920 --- /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-it", + "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..402067bda5e5bb88750932894d540dbd1bc4e9cb --- /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-it", + "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..31dea821f488507d12a58ae1092b58c32945b1b9 --- /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-it", + "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..31dea821f488507d12a58ae1092b58c32945b1b9 --- /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-it", + "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..edfb70326bb6c2a56bc361d33d0680b8d6a03322 --- /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-it", + "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..785c3fbce60424bcb8db286d556ae27263c76ce5 --- /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-it", + "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..55aaeb80f6164df2e0bf3388b4f7f2bfdb609954 --- /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-it", + "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..d007fc41f6d997bce5a66577824a8741c7ff19be --- /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-it", + "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..4b2763b9f9d20ba119000f66a6be024d7f030fdd --- /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-it", + "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..46dfdb44b3ad3d1ae90645edbd4265c20f88eef6 --- /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-it", + "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..a7afb89694c0ad4408513d4c47a0219f36a288c8 --- /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-it", + "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..32cd957267ffa0a360ff95d4dd3619ab118c77e1 --- /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-it", + "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..79682a319ea7f96583ff4f44653e6f5f04b5738e --- /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-it", + "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..7e42c0e103be9bc459511689a8270eb903dc3e6e --- /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-it", + "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..5d18e21d15a60db3efdc0224595a50fdbb3398f9 --- /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-it", + "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..6d90861f88041245668e920153151e8a96e68546 --- /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-it", + "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..6d90861f88041245668e920153151e8a96e68546 --- /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-it", + "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..1784681b5c5383ee3e0ea37acd011e1bd5329dd9 --- /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-it", + "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..7ce74b7c83664554e1936e3f7018899b799f06ac --- /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-it", + "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..17e599b82087479cb2d649d13ad228129e8d6d93 --- /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-it", + "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..cf79f9fc124917a09e61cf04329c0083fd04af38 --- /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-it", + "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..7b1510b6d8300833cb2c645048a0eadb5db3edd7 --- /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-it", + "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..c65dbc3be13bdcf328cdbe83d17b28c2d9afe141 --- /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-it", + "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..dc564b39bbcdced71d252447613c48d267a7810e --- /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-it", + "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..bb8d3bc66ffc67619031fce85fc837b33e412ea9 --- /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-it", + "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..ee277683b775dcc690bf35526cd9b2483b4cf48b --- /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-it", + "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..786e883eb0b6f1d9b773a985f5085a8641d4fbeb --- /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-it", + "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..2e24dad179e8f93e447c227a26cae5f45dbcad86 --- /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-it", + "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..915eb4d97029b71cb8595f94ef3d077f1aa934a4 --- /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-it", + "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..915eb4d97029b71cb8595f94ef3d077f1aa934a4 --- /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-it", + "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..194d500026bbd9f11659abae4e55a7d88dc66121 --- /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-it", + "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..216093c699c970ce5ba8a2b985aaf7d4aacd3ff1 --- /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-it", + "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..4ff3a63c6aa753f443316df70b6e3187e03fc9d9 --- /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-it", + "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..52c2877c1f5654bef212da837ac4decb320a13ca --- /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-it", + "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..39211eb607165857ba1b3b601997c3349ff2a470 --- /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-it", + "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..689ad906a35a6f6c5ef521b8f5b577d9c83f25d6 --- /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-it", + "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..0ce09979536a93a624d3a08e236e501ade40a241 --- /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-it", + "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..08c77028d369bfc6e8e890dda14085097091fa86 --- /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-it", + "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..23cb7fc6b296dd8e08bcd7b92fb1eeb3928fa2ed --- /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-it", + "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..9d9beae8b5747f01e00b88cd8fe485c32855951f --- /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-it", + "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..d98cac4a54524fb1f80fd80d5ad3d8aee3816541 --- /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-it", + "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..c84bdf94f21d49231a23ee9189c7f5c2295f750f --- /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-it", + "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..c84bdf94f21d49231a23ee9189c7f5c2295f750f --- /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-it", + "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..0239d135f60a1a6e63e30fcd683224a7378c67a3 --- /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-it", + "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..930acfef07df024f388b6287c15ec3fea6ee4a89 --- /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-it", + "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..05417ed7572f02df2718f11b5eaecd671a32f990 --- /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-it", + "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..49972ad186969651da139883e2361b58a522823e --- /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-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file