diff --git a/attn_out/layer_12_width_16k_l0_big/params.safetensors b/attn_out/layer_12_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e63d886acd943102072307d4ad3215a23a494c7f --- /dev/null +++ b/attn_out/layer_12_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f28368da006346daa050858f201ad31c77d007e2423b9ee161ff0ac564754e6 +size 134353272 diff --git a/attn_out/layer_12_width_16k_l0_medium/params.safetensors b/attn_out/layer_12_width_16k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..660a01ef21efdbfe4933c11880eb42d0a29804ba --- /dev/null +++ b/attn_out/layer_12_width_16k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e26d7eabfb7b586a6e8388e61c92e8672860e0200158ee1eafeb8133cb50cde0 +size 134353272 diff --git a/attn_out/layer_12_width_16k_l0_small/params.safetensors b/attn_out/layer_12_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..df13489031763cdf6b9217073d068ed979919283 --- /dev/null +++ b/attn_out/layer_12_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a115f88a1ab3f5eae9607373c5812f6e7fff1c02997626f4b0de8999ef053f4 +size 134353272 diff --git a/attn_out/layer_12_width_65k_l0_big/params.safetensors b/attn_out/layer_12_width_65k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cc618cf86579fba52939b6a4620cc971a378747f --- /dev/null +++ b/attn_out/layer_12_width_65k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb09a673f518aae52f1dbac67ac60ec446c29f6a716ff63da169c13a76b3abc5 +size 537399680 diff --git a/attn_out/layer_12_width_65k_l0_medium/params.safetensors b/attn_out/layer_12_width_65k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b29e8354e76bc83f21621f70bc0f4a6e4ba50804 --- /dev/null +++ b/attn_out/layer_12_width_65k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ffb322f341adf677569cdd32625c8069f99545cc1094c68fae97750da3bd450 +size 537399680 diff --git a/attn_out/layer_12_width_65k_l0_small/params.safetensors b/attn_out/layer_12_width_65k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0bb9a7049feb80f6717a1593c6b88594dd89b145 --- /dev/null +++ b/attn_out/layer_12_width_65k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c238ca353cb1aacba61f521688801cbd088ae6fb28398374ddc2d409e09f49e0 +size 537399680 diff --git a/attn_out/layer_15_width_16k_l0_big/params.safetensors b/attn_out/layer_15_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..08139d9e4016a95157f2d3565124a5c3883b7540 --- /dev/null +++ b/attn_out/layer_15_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcc0871ad39152806052828251c639d2fd4d2d3aeb7e58beb0c9fe27356fe66b +size 134353272 diff --git a/attn_out/layer_15_width_16k_l0_medium/params.safetensors b/attn_out/layer_15_width_16k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8329031b3481ccec29dca14739850d0070af3af0 --- /dev/null +++ b/attn_out/layer_15_width_16k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba7e101b9ac5ddf9044ece3bd24c48dbb1d59cec916fd282af7128a1df40a9ae +size 134353272 diff --git a/attn_out/layer_15_width_16k_l0_small/params.safetensors b/attn_out/layer_15_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e5aacf5eefc588b217110c8408c47d3e51ea0641 --- /dev/null +++ b/attn_out/layer_15_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a482def41c378a7324db6b4fa727691564985659ee3829781583f8483764abdb +size 134353272 diff --git a/attn_out/layer_15_width_65k_l0_big/params.safetensors b/attn_out/layer_15_width_65k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2f6b47a0da5b3777765925a3063734817702711a --- /dev/null +++ b/attn_out/layer_15_width_65k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0b4805b719e05259773a203d5495b3db2303beafafb77458bacb8b0f0d53ade +size 537399680 diff --git a/attn_out/layer_15_width_65k_l0_medium/params.safetensors b/attn_out/layer_15_width_65k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9e6d02e9c28b7b9ea7c447865efe831ea27377c7 --- /dev/null +++ b/attn_out/layer_15_width_65k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69125af286a2877fb612127d30ed08d7a93e351a624697368ce59e178cc2610e +size 537399680 diff --git a/attn_out/layer_15_width_65k_l0_small/params.safetensors b/attn_out/layer_15_width_65k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..273d920ac431ea01282d3ac03b1dc7c4abc29ce2 --- /dev/null +++ b/attn_out/layer_15_width_65k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3836a6b3c60fbbcb37b63b0caa894f0d0d064b2b2823825481bff069ba31e170 +size 537399680 diff --git a/attn_out/layer_5_width_16k_l0_big/params.safetensors b/attn_out/layer_5_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2f21e3640ce9a63c839c1848192185912face96f --- /dev/null +++ b/attn_out/layer_5_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d465f79c10c66ccf0f00d17b94efddb3d15237283a746d43a5b4623ea7ab9e2d +size 134353272 diff --git a/attn_out/layer_5_width_16k_l0_medium/params.safetensors b/attn_out/layer_5_width_16k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5c9600a57e9d43bb415f2d1ae0a847eff5507ece --- /dev/null +++ b/attn_out/layer_5_width_16k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13d2be8b5d9532cd7e061a953e281081118f97c48328ce2418ea866950b0aa9d +size 134353272 diff --git a/attn_out/layer_5_width_16k_l0_small/params.safetensors b/attn_out/layer_5_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5104172ed4132069cc25941403aee554cddffdb7 --- /dev/null +++ b/attn_out/layer_5_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:317372bb845c53222b98faea3c2fd0ec82c11c85752a6b1f9992336f60744fba +size 134353272 diff --git a/attn_out/layer_5_width_65k_l0_big/params.safetensors b/attn_out/layer_5_width_65k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d4cb5b9516b54fcce9f9d9599b5ca6c6b419134a --- /dev/null +++ b/attn_out/layer_5_width_65k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:709ec73b8009581f89471a73a26953eaeb4f576978b7d780b01b35055c435b70 +size 537399680 diff --git a/attn_out/layer_5_width_65k_l0_medium/params.safetensors b/attn_out/layer_5_width_65k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3e638d80773f5b409268bcf28c9619250dd35b8 --- /dev/null +++ b/attn_out/layer_5_width_65k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6873e146162b7bded50793313a8ebd63557a87be4c90b5a9cbe6889f9b472810 +size 537399680 diff --git a/attn_out/layer_5_width_65k_l0_small/params.safetensors b/attn_out/layer_5_width_65k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..75e05d08fe88122f873c20d15fddc5e660e4ee63 --- /dev/null +++ b/attn_out/layer_5_width_65k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8cfa565a59b59dbdaf0020944f676aa5d9d7f4f18fa91a8f730de476d3857a03 +size 537399680 diff --git a/attn_out/layer_9_width_16k_l0_big/params.safetensors b/attn_out/layer_9_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f5b61246e25cf8c7c8ca7211853246880e2e9051 --- /dev/null +++ b/attn_out/layer_9_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a602e7379fef1b8ca6d9de5f71afe1dc701fbed21734ef239f92c9d7027d8fcb +size 134353272 diff --git a/attn_out/layer_9_width_16k_l0_medium/params.safetensors b/attn_out/layer_9_width_16k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..336c12a0aa513fdd213d474c71a9861d96e8bb5b --- /dev/null +++ b/attn_out/layer_9_width_16k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c4cc52c98dcb24494df4345b130f9443f49dcfa4b62745f2258f5ca3b28072c +size 134353272 diff --git a/attn_out/layer_9_width_16k_l0_small/params.safetensors b/attn_out/layer_9_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..346f1acea2935d56c6b2eb9179cdc9dc0d8ddd8e --- /dev/null +++ b/attn_out/layer_9_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:978eea8ed95c313def1b5c0e20885ef4886b84548c957a9f61a5eebd872e2020 +size 134353272 diff --git a/attn_out/layer_9_width_65k_l0_big/params.safetensors b/attn_out/layer_9_width_65k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cc4abf59a706f16ff68f4d4fe8e84cf44e391dd2 --- /dev/null +++ b/attn_out/layer_9_width_65k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f591fb8142312fced18512af06b4120d9b269d3f44e118d671f623041e43536d +size 537399680 diff --git a/attn_out/layer_9_width_65k_l0_medium/params.safetensors b/attn_out/layer_9_width_65k_l0_medium/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..291d7826158032394a4210fcf3967c49fe402558 --- /dev/null +++ b/attn_out/layer_9_width_65k_l0_medium/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06f90ff0d7c44f2ecef5bc77aa48a4616c7e987d52e543ea83a08b0925dc79ad +size 537399680 diff --git a/attn_out/layer_9_width_65k_l0_small/params.safetensors b/attn_out/layer_9_width_65k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..61587c8136694c8b78a5f48bb3302905aaa346af --- /dev/null +++ b/attn_out/layer_9_width_65k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a420e72d57e3130cba1827fe6b576e4e2f18b30fd76445effcb1a78c8d475609 +size 537399680 diff --git a/resid_post_all/layer_0_width_16k_l0_big/config.json b/resid_post_all/layer_0_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3940136bb94165d3eb5962ce2de30cca10918e3d --- /dev/null +++ b/resid_post_all/layer_0_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.output", + "hf_hook_point_out": "model.layers.0.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_all/layer_0_width_16k_l0_small/config.json b/resid_post_all/layer_0_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f1764c489e1abcf86a0ed7d0ae3952fad8b1bd58 --- /dev/null +++ b/resid_post_all/layer_0_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.output", + "hf_hook_point_out": "model.layers.0.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_0_width_262k_l0_big/config.json b/resid_post_all/layer_0_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b0cd413a43caaae246619f21ebac3d2966a89e17 --- /dev/null +++ b/resid_post_all/layer_0_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.output", + "hf_hook_point_out": "model.layers.0.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_all/layer_0_width_262k_l0_small/config.json b/resid_post_all/layer_0_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6253f9636319afed30135bafe948c6f77c34269d --- /dev/null +++ b/resid_post_all/layer_0_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.0.output", + "hf_hook_point_out": "model.layers.0.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 10, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_10_width_16k_l0_small/config.json b/resid_post_all/layer_10_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dcd7e6fe8ff5f2021984c2386bdba1fb71193ccd --- /dev/null +++ b/resid_post_all/layer_10_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.output", + "hf_hook_point_out": "model.layers.10.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_10_width_262k_l0_big/config.json b/resid_post_all/layer_10_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5fe2a02f97c5f1b20efe6f08bd9f1e8eb47ef7da --- /dev/null +++ b/resid_post_all/layer_10_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.output", + "hf_hook_point_out": "model.layers.10.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_10_width_262k_l0_small/config.json b/resid_post_all/layer_10_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ba642b19628a05a4335b7f2032ad634390a9f65f --- /dev/null +++ b/resid_post_all/layer_10_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.10.output", + "hf_hook_point_out": "model.layers.10.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_11_width_16k_l0_big/config.json b/resid_post_all/layer_11_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5002ddeab0d9715cc7c492e9025b5ea66b2bcaf0 --- /dev/null +++ b/resid_post_all/layer_11_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.output", + "hf_hook_point_out": "model.layers.11.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_11_width_16k_l0_small/config.json b/resid_post_all/layer_11_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6a6976430ef8b121237857a5d281c92b382de68b --- /dev/null +++ b/resid_post_all/layer_11_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.output", + "hf_hook_point_out": "model.layers.11.output", + "width": 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_all/layer_11_width_262k_l0_big/config.json b/resid_post_all/layer_11_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a2fa08b829fed905aa2a2852aba8f19f5d1b131b --- /dev/null +++ b/resid_post_all/layer_11_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.11.output", + "hf_hook_point_out": "model.layers.11.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_12_width_16k_l0_big/config.json b/resid_post_all/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..721bcd15f839f533649c3d39a9cd6736da386a41 --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_12_width_16k_l0_small/config.json b/resid_post_all/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..44fb5bb602ef875ca5c62c99117148ed6f4d54ea --- /dev/null +++ b/resid_post_all/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_all/layer_12_width_262k_l0_big/config.json b/resid_post_all/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6490318f53e3c5be128d69b08a85d15069a256bc --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_12_width_262k_l0_small/config.json b/resid_post_all/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..edfb70326bb6c2a56bc361d33d0680b8d6a03322 --- /dev/null +++ b/resid_post_all/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_all/layer_13_width_16k_l0_big/config.json b/resid_post_all/layer_13_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a935f96bb486b569df40f10cad8853cecdb6f27a --- /dev/null +++ b/resid_post_all/layer_13_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.output", + "hf_hook_point_out": "model.layers.13.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_13_width_16k_l0_small/config.json b/resid_post_all/layer_13_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c000ccebfb6f7ab8dab11a5371dd9301d8410ff5 --- /dev/null +++ b/resid_post_all/layer_13_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.output", + "hf_hook_point_out": "model.layers.13.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_all/layer_13_width_262k_l0_big/config.json b/resid_post_all/layer_13_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0d92b7feb1f903de485e736ba9075ab41bed8806 --- /dev/null +++ b/resid_post_all/layer_13_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.output", + "hf_hook_point_out": "model.layers.13.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_13_width_262k_l0_small/config.json b/resid_post_all/layer_13_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ac8b13d04056d5d8eb1bda0bd029b890e4961149 --- /dev/null +++ b/resid_post_all/layer_13_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.13.output", + "hf_hook_point_out": "model.layers.13.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_14_width_16k_l0_big/config.json b/resid_post_all/layer_14_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ef4e8c96a0ebe3fe989be817ee99c0d57abd81aa --- /dev/null +++ b/resid_post_all/layer_14_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.output", + "hf_hook_point_out": "model.layers.14.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_14_width_16k_l0_small/config.json b/resid_post_all/layer_14_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5ed0cc7cd7677382f010f775858d7a2937cb8e88 --- /dev/null +++ b/resid_post_all/layer_14_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.output", + "hf_hook_point_out": "model.layers.14.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_all/layer_14_width_262k_l0_big/config.json b/resid_post_all/layer_14_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fc1d9a94355d0b33bb1ba65c3d2bfb48c1ac8571 --- /dev/null +++ b/resid_post_all/layer_14_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.output", + "hf_hook_point_out": "model.layers.14.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_14_width_262k_l0_small/config.json b/resid_post_all/layer_14_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..638f93748a87dc67c44d47d7d9d925d650391413 --- /dev/null +++ b/resid_post_all/layer_14_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.14.output", + "hf_hook_point_out": "model.layers.14.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_15_width_16k_l0_big/config.json b/resid_post_all/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8d42f95ec81a80c049b7a202f2e487c1eeb9e5b5 --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_15_width_16k_l0_small/config.json b/resid_post_all/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7afb89694c0ad4408513d4c47a0219f36a288c8 --- /dev/null +++ b/resid_post_all/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_all/layer_15_width_262k_l0_big/config.json b/resid_post_all/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a2b6c8ac18b144702749758159920ef24c1cee5b --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_15_width_262k_l0_small/config.json b/resid_post_all/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1784681b5c5383ee3e0ea37acd011e1bd5329dd9 --- /dev/null +++ b/resid_post_all/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_all/layer_16_width_16k_l0_big/config.json b/resid_post_all/layer_16_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5f861c7201ebf840974e849a12a17112eb94dd02 --- /dev/null +++ b/resid_post_all/layer_16_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.output", + "hf_hook_point_out": "model.layers.16.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_16_width_16k_l0_small/config.json b/resid_post_all/layer_16_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c2a5397d05ff85a5a87e2b5976b82300b5e43518 --- /dev/null +++ b/resid_post_all/layer_16_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.output", + "hf_hook_point_out": "model.layers.16.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_all/layer_16_width_262k_l0_big/config.json b/resid_post_all/layer_16_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e71a5b6f33fb75eadc98407b44fe4cd4cef4ff94 --- /dev/null +++ b/resid_post_all/layer_16_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.output", + "hf_hook_point_out": "model.layers.16.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_16_width_262k_l0_small/config.json b/resid_post_all/layer_16_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a5e29a6426ab702d30d2a5d4175eb6ead25fcae3 --- /dev/null +++ b/resid_post_all/layer_16_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.16.output", + "hf_hook_point_out": "model.layers.16.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_17_width_16k_l0_small/config.json b/resid_post_all/layer_17_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fbab0773b9a21602af6d9163864335a55ccf9145 --- /dev/null +++ b/resid_post_all/layer_17_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_17_width_262k_l0_big/config.json b/resid_post_all/layer_17_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e1a09bef0d13fb7571e9010f8a8b92f76546d74d --- /dev/null +++ b/resid_post_all/layer_17_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_17_width_262k_l0_small/config.json b/resid_post_all/layer_17_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5339bccfc60ea01d4435ca271be33ab46ccb975e --- /dev/null +++ b/resid_post_all/layer_17_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.17.output", + "hf_hook_point_out": "model.layers.17.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_1_width_16k_l0_big/config.json b/resid_post_all/layer_1_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..300ca1fd06da09bd348563fde3c9a3df3becd92a --- /dev/null +++ b/resid_post_all/layer_1_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.output", + "hf_hook_point_out": "model.layers.1.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_1_width_16k_l0_small/config.json b/resid_post_all/layer_1_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..65a0c157d842b0add61d083cd95800b7acdd48d1 --- /dev/null +++ b/resid_post_all/layer_1_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.output", + "hf_hook_point_out": "model.layers.1.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_1_width_262k_l0_big/config.json b/resid_post_all/layer_1_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..65aaf2eb7a013e77665e248bd8c5f06b2fd04b21 --- /dev/null +++ b/resid_post_all/layer_1_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.output", + "hf_hook_point_out": "model.layers.1.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 70, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_1_width_262k_l0_small/config.json b/resid_post_all/layer_1_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..53a2999a6f4a073d1052bc66e106ce5a7329f02c --- /dev/null +++ b/resid_post_all/layer_1_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.1.output", + "hf_hook_point_out": "model.layers.1.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 11, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_2_width_16k_l0_big/config.json b/resid_post_all/layer_2_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..22df56fa41b4f243d1e02238aa9069e3823b89eb --- /dev/null +++ b/resid_post_all/layer_2_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.output", + "hf_hook_point_out": "model.layers.2.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 80, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_2_width_16k_l0_small/config.json b/resid_post_all/layer_2_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7060e33354560d4cf30edf5a688f203c871eb0af --- /dev/null +++ b/resid_post_all/layer_2_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.2.output", + "hf_hook_point_out": "model.layers.2.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 13, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_3_width_16k_l0_big/config.json b/resid_post_all/layer_3_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e315ce2b60bcc27a51558f70a7bc6a5c54fe02eb --- /dev/null +++ b/resid_post_all/layer_3_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.output", + "hf_hook_point_out": "model.layers.3.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_3_width_16k_l0_small/config.json b/resid_post_all/layer_3_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cfa325648042a301648dfd6cceb832507dd6df --- /dev/null +++ b/resid_post_all/layer_3_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.output", + "hf_hook_point_out": "model.layers.3.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_3_width_262k_l0_big/config.json b/resid_post_all/layer_3_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..240f77cd186381bdebbf2e620a6d97128ed356ef --- /dev/null +++ b/resid_post_all/layer_3_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.output", + "hf_hook_point_out": "model.layers.3.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 90, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_3_width_262k_l0_small/config.json b/resid_post_all/layer_3_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..43e2676a92ec063d39c654c2203647f9ad5beb0b --- /dev/null +++ b/resid_post_all/layer_3_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.3.output", + "hf_hook_point_out": "model.layers.3.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 15, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_4_width_16k_l0_big/config.json b/resid_post_all/layer_4_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db7fc3581d61b24a4d26a6662bf8d97759058d66 --- /dev/null +++ b/resid_post_all/layer_4_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.output", + "hf_hook_point_out": "model.layers.4.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 100, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_4_width_16k_l0_small/config.json b/resid_post_all/layer_4_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..68980770f4716c88ca848befafbe70a051d76465 --- /dev/null +++ b/resid_post_all/layer_4_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.output", + "hf_hook_point_out": "model.layers.4.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_4_width_262k_l0_big/config.json b/resid_post_all/layer_4_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d8755cddf688ab56da2a2a9d6de52722d3d16ff2 --- /dev/null +++ b/resid_post_all/layer_4_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.output", + "hf_hook_point_out": "model.layers.4.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 100, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_4_width_262k_l0_small/config.json b/resid_post_all/layer_4_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e095f626f5e7cd1b49eb757ac155e8ffd6911353 --- /dev/null +++ b/resid_post_all/layer_4_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.4.output", + "hf_hook_point_out": "model.layers.4.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 16, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_5_width_16k_l0_big/config.json b/resid_post_all/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7d414023c795fc71e608350a952d081d565d9dd8 --- /dev/null +++ b/resid_post_all/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": 110, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_5_width_16k_l0_small/config.json b/resid_post_all/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dc564b39bbcdced71d252447613c48d267a7810e --- /dev/null +++ b/resid_post_all/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_all/layer_5_width_262k_l0_big/config.json b/resid_post_all/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a6c0a20fc870b3b30137a16678c54c85619b505 --- /dev/null +++ b/resid_post_all/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": 110, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_5_width_262k_l0_small/config.json b/resid_post_all/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..194d500026bbd9f11659abae4e55a7d88dc66121 --- /dev/null +++ b/resid_post_all/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_all/layer_6_width_16k_l0_big/config.json b/resid_post_all/layer_6_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..613e373b685982b526641178c465d819cefb57c7 --- /dev/null +++ b/resid_post_all/layer_6_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.output", + "hf_hook_point_out": "model.layers.6.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_6_width_16k_l0_small/config.json b/resid_post_all/layer_6_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3207b9c1d0fa2ffbffacbd2b77698f9a28a8e714 --- /dev/null +++ b/resid_post_all/layer_6_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.output", + "hf_hook_point_out": "model.layers.6.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_all/layer_6_width_262k_l0_big/config.json b/resid_post_all/layer_6_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0ff04a2ff6075cf0206b2bbe33bf699ef97514a5 --- /dev/null +++ b/resid_post_all/layer_6_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.output", + "hf_hook_point_out": "model.layers.6.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_6_width_262k_l0_small/config.json b/resid_post_all/layer_6_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..30363f8e504201b67de85adacf7e06be87b3cc16 --- /dev/null +++ b/resid_post_all/layer_6_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.6.output", + "hf_hook_point_out": "model.layers.6.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_7_width_16k_l0_big/config.json b/resid_post_all/layer_7_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1dbe95fc73ce2662e441faa762fe45682dd3bc56 --- /dev/null +++ b/resid_post_all/layer_7_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.output", + "hf_hook_point_out": "model.layers.7.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_7_width_16k_l0_small/config.json b/resid_post_all/layer_7_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6435c8cd7309d2b96fc12eef6fa18263e15631c8 --- /dev/null +++ b/resid_post_all/layer_7_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.output", + "hf_hook_point_out": "model.layers.7.output", + "width": 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_all/layer_7_width_262k_l0_big/config.json b/resid_post_all/layer_7_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..083dda38bee387766d565f9c92369e8473fa871c --- /dev/null +++ b/resid_post_all/layer_7_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.7.output", + "hf_hook_point_out": "model.layers.7.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_8_width_16k_l0_big/config.json b/resid_post_all/layer_8_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e04022d58c5a3d06a0859e90cec92f9237e2c108 --- /dev/null +++ b/resid_post_all/layer_8_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.output", + "hf_hook_point_out": "model.layers.8.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_8_width_16k_l0_small/config.json b/resid_post_all/layer_8_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db58ef049308ca36377b45e645c02e95373ca062 --- /dev/null +++ b/resid_post_all/layer_8_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.output", + "hf_hook_point_out": "model.layers.8.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_all/layer_8_width_262k_l0_big/config.json b/resid_post_all/layer_8_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9233160e513bbd46565e037d3ac5284b83376887 --- /dev/null +++ b/resid_post_all/layer_8_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.output", + "hf_hook_point_out": "model.layers.8.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_8_width_262k_l0_small/config.json b/resid_post_all/layer_8_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..14f9015b292262207b085ccfbf9b13d347172a41 --- /dev/null +++ b/resid_post_all/layer_8_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.8.output", + "hf_hook_point_out": "model.layers.8.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_9_width_16k_l0_big/config.json b/resid_post_all/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a6d31517cfbd1b4aa86339096ae2f18d7630382c --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_9_width_262k_l0_big/config.json b/resid_post_all/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ab2341e3afc44bd4bb6c58bd40cc51d0e75485df --- /dev/null +++ b/resid_post_all/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": 120, + "affine_connection": false +} \ No newline at end of file diff --git a/resid_post_all/layer_9_width_262k_l0_small/config.json b/resid_post_all/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0239d135f60a1a6e63e30fcd683224a7378c67a3 --- /dev/null +++ b/resid_post_all/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/transcoder/layer_12_width_16k_l0_big/config.json b/transcoder/layer_12_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f52be52f05ae204457770c790f9e556da7cd0cae --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_big_affine/config.json b/transcoder/layer_12_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..85b04ca06230fe04e71e4f23fb4c6069d960bbe2 --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_medium/config.json b/transcoder/layer_12_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..86e81cfa5e7c1d1db3e49bb48516cb501e56868e --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_medium_affine/config.json b/transcoder/layer_12_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aaf0f272de1232a5672be37944a521a69d0ba6be --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_small/config.json b/transcoder/layer_12_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6b209a7dc5801476fc329713e47340e7fc1786de --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_16k_l0_small_affine/config.json b/transcoder/layer_12_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c2cfe43632fc0bfe3e8472f261911cc961bf71bc --- /dev/null +++ b/transcoder/layer_12_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_big/config.json b/transcoder/layer_12_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ebd6f114e76b59769a240aaeccfe27f99f4afea --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_big_affine/config.json b/transcoder/layer_12_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9aff9c20051020fa49c334ff915dc8dfda29d65a --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_medium/config.json b/transcoder/layer_12_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..77c4a0b01998687819d6a3b0cf3d1a52aec41563 --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_medium_affine/config.json b/transcoder/layer_12_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea413282b8fdb377b03d76cddec2e62d0fa3de88 --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_small/config.json b/transcoder/layer_12_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..189998bda15ccb0488124214e0ff092ff5a52110 --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_262k_l0_small_affine/config.json b/transcoder/layer_12_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..65f98cb9c2afd319d63f878804d4160bb116a68d --- /dev/null +++ b/transcoder/layer_12_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_big/config.json b/transcoder/layer_12_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12815bf074560d0898193080de12a6a1823cbd89 --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_big_affine/config.json b/transcoder/layer_12_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2ead86fb409dd097cda6f0dcc65330d21da851e1 --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_medium/config.json b/transcoder/layer_12_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..80f930bc141c1cb61de56af84866873dd5f2c18e --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_medium_affine/config.json b/transcoder/layer_12_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..53a1a8a4377ad68f80950b8f7e15eaa804c971eb --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_small/config.json b/transcoder/layer_12_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7ed4e2afc2dc5fa1481d7eafc0c8ff16eebdc78d --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_12_width_65k_l0_small_affine/config.json b/transcoder/layer_12_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f1e1ceeebccf8df69126f74fb68d1511340170d7 --- /dev/null +++ b/transcoder/layer_12_width_65k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.12.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.12.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_big/config.json b/transcoder/layer_15_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..154cdb5c7a3ec1ca7b1fb476ca37adb23c8c8a84 --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_big_affine/config.json b/transcoder/layer_15_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b2b606a7602542a019e3c7f1fa0dcd31e3d775a --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_medium/config.json b/transcoder/layer_15_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..97115cdd05a990428c09a545e62e0c3672d882ef --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_medium_affine/config.json b/transcoder/layer_15_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6ef648d94cc1efead1dbc32caf859dc50a05e90f --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_small/config.json b/transcoder/layer_15_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7d7dbcfe4f01649b82237bc5e43958b4da566e7c --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_16k_l0_small_affine/config.json b/transcoder/layer_15_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07393a25faf514c5397b06041d786d719739acd3 --- /dev/null +++ b/transcoder/layer_15_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_big/config.json b/transcoder/layer_15_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d659f2bdbb760194911a4769501c8aa760696b8a --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_big_affine/config.json b/transcoder/layer_15_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4fdc9b782b0b8793d7a3408829d86ea1e4785e5 --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_medium/config.json b/transcoder/layer_15_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..02b30cf19bee9a5a41915e7f8a48a2baf28bd0b2 --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_medium_affine/config.json b/transcoder/layer_15_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..15d30f8c8b5b414bde3e0f301c4c5e79e930e35d --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_small/config.json b/transcoder/layer_15_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..460a1dd6fa92d6430e40f1502d8f607a1e7b6538 --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_262k_l0_small_affine/config.json b/transcoder/layer_15_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e4f93daff828fa013371262b6499c2ace4f380d7 --- /dev/null +++ b/transcoder/layer_15_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_big/config.json b/transcoder/layer_15_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..72f0521575ca36cea91f054f9d953fe8af6f458c --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_big_affine/config.json b/transcoder/layer_15_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1f993289e2337f95ec1602ad21660a62b036f69d --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_medium/config.json b/transcoder/layer_15_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8ce36e590674d3d5b6e7872bb3ec6bee45ede52d --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_medium_affine/config.json b/transcoder/layer_15_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..25facd471ea965951706218525957d6b79285486 --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_small/config.json b/transcoder/layer_15_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..669766bcd3c7d8a4b86d723de3750bc5e31a1f80 --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_15_width_65k_l0_small_affine/config.json b/transcoder/layer_15_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..562f741b1080f4b3e2ba70ea13c7f7580f09f7c9 --- /dev/null +++ b/transcoder/layer_15_width_65k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.15.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.15.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_big/config.json b/transcoder/layer_5_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b3314d5b35ec87a6c148ac12f5c32d097d25b1c0 --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_big_affine/config.json b/transcoder/layer_5_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..701051fabc6046a4e6cd4628b776d825e0380ea0 --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_medium/config.json b/transcoder/layer_5_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6bf156dfc5335f13ff3e4613f0a7d0115f3e5011 --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_medium_affine/config.json b/transcoder/layer_5_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cd084a050487bc705ea9217dc27848ffe86fe0f7 --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_small/config.json b/transcoder/layer_5_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..433eb32b49037ba116932e436aff6444bc560293 --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_16k_l0_small_affine/config.json b/transcoder/layer_5_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4455a1472ec311214577a5501991bc225fa36e3e --- /dev/null +++ b/transcoder/layer_5_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_big/config.json b/transcoder/layer_5_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f0ca583bad552551ed7033684a3ea0d282928d69 --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_big_affine/config.json b/transcoder/layer_5_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..736998665766fd11d1ed93f119b8fa4038f1a565 --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_medium/config.json b/transcoder/layer_5_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7951e7d1c025c0ebb9d0e455ea69644ee47b0362 --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_medium_affine/config.json b/transcoder/layer_5_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c9d0ea15cdae62926fe8ffef075f78d0d9e5cb1e --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_small/config.json b/transcoder/layer_5_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..51a308ea45bf2d7f922722a504eec8ec275daad7 --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_262k_l0_small_affine/config.json b/transcoder/layer_5_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc25118bd55a2b71e9c50ca8ffe9272520ef2d72 --- /dev/null +++ b/transcoder/layer_5_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_big/config.json b/transcoder/layer_5_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e021ab18957ea18e08a5a04cfc720691fa4aac9 --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_big_affine/config.json b/transcoder/layer_5_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8cb7afbec5ca99c543597d9fed9e21d3a34ff990 --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 135, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_medium/config.json b/transcoder/layer_5_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f3a0116587a4aa035db5cb8001701126659b24e7 --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_medium_affine/config.json b/transcoder/layer_5_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..28bc6cd62c4bc6458e80657b311f313aa40a12a2 --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 55, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_small/config.json b/transcoder/layer_5_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c0cd95653055ddaef4423976171ff4630987128b --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_5_width_65k_l0_small_affine/config.json b/transcoder/layer_5_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5f1e17c746bdb571d8239476c89a7c07ee97a964 --- /dev/null +++ b/transcoder/layer_5_width_65k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.5.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.5.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 18, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_big/config.json b/transcoder/layer_9_width_16k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..41219043bc7746a461ee941b8d8c9e63100996cc --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_big_affine/config.json b/transcoder/layer_9_width_16k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..963ddc4d98a6f406cc05f0fb9a2286a16d62e61b --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium/config.json b/transcoder/layer_9_width_16k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..10b12538400353299c1e38567f63ed91cb6f1521 --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_medium_affine/config.json b/transcoder/layer_9_width_16k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d65d9d4bf3b4ec6a1afead199b77825a391f11e1 --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_small/config.json b/transcoder/layer_9_width_16k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..84836a0a310a8691331512b48d1bc6f689925012 --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_16k_l0_small_affine/config.json b/transcoder/layer_9_width_16k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ea0330788c25238d6244aa55b4c0a08f9389aab --- /dev/null +++ b/transcoder/layer_9_width_16k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 16384, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big/config.json b/transcoder/layer_9_width_262k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a53dbc2cd9fcb3abbe1abda1ee17d30f1046df5c --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_big_affine/config.json b/transcoder/layer_9_width_262k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b693df65bab2802db7ac16f4278881c2ed31932b --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium/config.json b/transcoder/layer_9_width_262k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..79767800bfa35ba24dcb39db3e152d3ec718a0f0 --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_medium_affine/config.json b/transcoder/layer_9_width_262k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..927afe21897c7ebdcf117213b4feb7948f68393b --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small/config.json b/transcoder/layer_9_width_262k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd00138ab3d7355f49cd535205f4cda4bd7fcfed --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_262k_l0_small_affine/config.json b/transcoder/layer_9_width_262k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d8e18e8d79d6ea5a843faa77f0d01b254da5549a --- /dev/null +++ b/transcoder/layer_9_width_262k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 262144, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_big/config.json b/transcoder/layer_9_width_65k_l0_big/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d64c6b970ae44fd70141651a1d5211e51d5075be --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_big/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_big_affine/config.json b/transcoder/layer_9_width_65k_l0_big_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..af3e88eba8083290e877807759d26f3bebf624e9 --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_big_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 150, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium/config.json b/transcoder/layer_9_width_65k_l0_medium/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d9e59c282aee967f8587828750247d16c0dcb6df --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_medium/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_medium_affine/config.json b/transcoder/layer_9_width_65k_l0_medium_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6b9bea1a337adc47fcd5bbc2da6a561601bef49e --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_medium_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 60, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small/config.json b/transcoder/layer_9_width_65k_l0_small/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6c4f386c3c9f1c08c64286e8e3786e6acbdd107a --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_small/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": false +} \ No newline at end of file diff --git a/transcoder/layer_9_width_65k_l0_small_affine/config.json b/transcoder/layer_9_width_65k_l0_small_affine/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aca6f4e52af848d117a9ed1c5854ab639b5f2081 --- /dev/null +++ b/transcoder/layer_9_width_65k_l0_small_affine/config.json @@ -0,0 +1,9 @@ +{ + "hf_hook_point_in": "model.layers.9.pre_feedforward_layernorm.output", + "hf_hook_point_out": "model.layers.9.post_feedforward_layernorm.output", + "width": 65536, + "model_name": "gemma-v3-270m-it", + "architecture": "jump_relu", + "l0": 20, + "affine_connection": true +} \ No newline at end of file diff --git a/transcoder_all/layer_0_width_16k_l0_big/params.safetensors b/transcoder_all/layer_0_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7f490d5af8e64b6fe4a53630f61ea142d8a12ff7 --- /dev/null +++ b/transcoder_all/layer_0_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e6775d7d4a9a72a0021d154c2180dc626a8219deb65ce25aea7e9d1f518f97bf +size 84020088 diff --git a/transcoder_all/layer_0_width_16k_l0_small/params.safetensors b/transcoder_all/layer_0_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..04d4d35666d4856a52c8fba4e821af60ac205d53 --- /dev/null +++ b/transcoder_all/layer_0_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:435070f2466f281b7a8ef69b803f5ed0c247aac69665cf754507ce9d0652dffd +size 84020088 diff --git a/transcoder_all/layer_0_width_262k_l0_big/params.safetensors b/transcoder_all/layer_0_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8abf96a13c46b8499b73d2687691a9d92becae25 --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f9506ad588839d95b706adf52e73783d3a4969f81946ec0c8e1059ab09cef379 +size 1344277376 diff --git a/transcoder_all/layer_0_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_0_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7f71ff37f92255d602bfb7ce9049f6a4e93e16aa --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e687feb91a6b56ffc8063fbe5ff07898d748d79356c8c806d4771f14bddf3fbe +size 1345915880 diff --git a/transcoder_all/layer_0_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_0_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..04f67972beb031821753be7fc7b49ab240450ddb --- /dev/null +++ b/transcoder_all/layer_0_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ee44aeb0dd35f9d54ca30d056a31bd83c6b2252eb09ec9f2b92b2c0b695e5a5 +size 1345915880 diff --git a/transcoder_all/layer_10_width_16k_l0_big/params.safetensors b/transcoder_all/layer_10_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b3e08161cb22f76e0f59487ae6753b695d7e7fc0 --- /dev/null +++ b/transcoder_all/layer_10_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d34b22211779ed3e8806a0953c85d543ad0cc0db94490c8a5f7aedf5a631af2 +size 84020088 diff --git a/transcoder_all/layer_10_width_262k_l0_big/params.safetensors b/transcoder_all/layer_10_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..767657ffe6f9e71507deec3b4b3e72d0b146d97e --- /dev/null +++ b/transcoder_all/layer_10_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd72f6e6252c48fd13da8dd96fdba21a4e65930f970b965569e806ee8468ad33 +size 1344277376 diff --git a/transcoder_all/layer_11_width_262k_l0_small/params.safetensors b/transcoder_all/layer_11_width_262k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..de915ebd7e67c1062c590b7dd5823dee01aded42 --- /dev/null +++ b/transcoder_all/layer_11_width_262k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:407d62d9ba77207c7cc42b592e2a1ee09e13f47947ff0482a3242cd9cd5f6177 +size 1344277376 diff --git a/transcoder_all/layer_12_width_16k_l0_big_affine/params.safetensors b/transcoder_all/layer_12_width_16k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a21bbb163b85690e564438225fccb64e19ba4897 --- /dev/null +++ b/transcoder_all/layer_12_width_16k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ada6b34028d97a7b3d309a415105f215559bcdfef6fc132277522bc809804e6 +size 85658592 diff --git a/transcoder_all/layer_12_width_16k_l0_small_affine/params.safetensors b/transcoder_all/layer_12_width_16k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..48c528fb18fea16da53379ddcb4e0405f5d3b8e3 --- /dev/null +++ b/transcoder_all/layer_12_width_16k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d57a73c7aac560eefb492dad0105314006f784abd0125a03f7e7c090c04cc85f +size 85658592 diff --git a/transcoder_all/layer_13_width_16k_l0_big/params.safetensors b/transcoder_all/layer_13_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2cd3a463fd4a40435e003ed6349c3e81246e6fd4 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2dd4389889912404527644fa0ff294f2d7e592fca94196831e77ddcc9b9f5d85 +size 84020088 diff --git a/transcoder_all/layer_13_width_16k_l0_big_affine/params.safetensors b/transcoder_all/layer_13_width_16k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..de31bf7ffdcfc7ef704e1137e94fcda95f127f63 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62ecdfc6da60f2d050071b814639c61f2f5e1e13ff30034887e7dd5f7a5e479f +size 85658592 diff --git a/transcoder_all/layer_13_width_16k_l0_small/params.safetensors b/transcoder_all/layer_13_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..70068c49878df1e9b94d57b8eaa3c5f2bec659b5 --- /dev/null +++ b/transcoder_all/layer_13_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8204ea2f65c6c90d8b10626bd1f2fb86208bcdbfa1c768648f17c01d10447511 +size 84020088 diff --git a/transcoder_all/layer_14_width_262k_l0_small/params.safetensors b/transcoder_all/layer_14_width_262k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..01a5c6072ad3a9ccd8308989afe861b094b159f5 --- /dev/null +++ b/transcoder_all/layer_14_width_262k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9cf28dd7b4ce6892dbad19076dad78521f84853c02224aae5ab0d326d4f9163f +size 1344277376 diff --git a/transcoder_all/layer_15_width_16k_l0_big/params.safetensors b/transcoder_all/layer_15_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e8e8dc01aebaadf31542e2148e537d1d05ceb665 --- /dev/null +++ b/transcoder_all/layer_15_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d03696c94b12afd8e8b0af72b155e44845d2ba295482971b8a999c826d6307a +size 84020088 diff --git a/transcoder_all/layer_15_width_262k_l0_big/params.safetensors b/transcoder_all/layer_15_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f5c70b4ad8bb2d164a4533900db639d6568a5c20 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ad14337a4ba0242b615eff4ee67787c38d3ff6e36f803bbbe92e91781b315c8 +size 1344277376 diff --git a/transcoder_all/layer_15_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_15_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cf07afce762846214b6951d00c3211048f78e093 --- /dev/null +++ b/transcoder_all/layer_15_width_262k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fa60d2dd681d0f77b0663ccc014659b680a2a498d48f12b5bc4bd6e62b6ee10 +size 1345915880 diff --git a/transcoder_all/layer_17_width_16k_l0_big/params.safetensors b/transcoder_all/layer_17_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5909b4bf54bdf5ea83532a408dbd2c83881d3fc9 --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:630b80ffe04b246025459419d724335f5f130004fc0b20d8c3d63363045b142f +size 84020088 diff --git a/transcoder_all/layer_17_width_16k_l0_big_affine/params.safetensors b/transcoder_all/layer_17_width_16k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..733712afe1d64618fb9568508decffd70b729bde --- /dev/null +++ b/transcoder_all/layer_17_width_16k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1d84027249dadd5fdf2ad1f65d96cef96be57c54e011490f109c67a99f31256 +size 85658592 diff --git a/transcoder_all/layer_17_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_17_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..231fb9d7dd32c52851824838d628d62cb4fc40d3 --- /dev/null +++ b/transcoder_all/layer_17_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c989b2d5473d63b652155e53c5587579af8a1298c9cca08c85677c7a14347d28 +size 1345915880 diff --git a/transcoder_all/layer_1_width_16k_l0_big_affine/params.safetensors b/transcoder_all/layer_1_width_16k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..35161290f7e1db65f5083013c826eb5033b8eb28 --- /dev/null +++ b/transcoder_all/layer_1_width_16k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ede2ab333fd01b28e2963c203bb8f0b29c1fb13c692d43faa2d50455118690f +size 85658592 diff --git a/transcoder_all/layer_1_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_1_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3b79983b1e41589104876b0c318f6c1098a29e8c --- /dev/null +++ b/transcoder_all/layer_1_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9ff73722e139c2ee15e7e2250a8540f3925ee9ab97edc92ee5a8e32ae549827 +size 1345915880 diff --git a/transcoder_all/layer_2_width_262k_l0_big/params.safetensors b/transcoder_all/layer_2_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..448f9917af804da6f1b6fe47576909fd37acdca2 --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37d2ebb1458de62805b56cb6b6a2164ecb393137ed40c72ae8ae9478a7ead5d2 +size 1344277376 diff --git a/transcoder_all/layer_2_width_262k_l0_small/params.safetensors b/transcoder_all/layer_2_width_262k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..72bf15c846ba49e575af27a3c1fdec4a128a021c --- /dev/null +++ b/transcoder_all/layer_2_width_262k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06d469bef0d6c117870c465e6857acc235f1546f44f02917ef7f6c991e47d86c +size 1344277376 diff --git a/transcoder_all/layer_3_width_16k_l0_small/params.safetensors b/transcoder_all/layer_3_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2fd94e7221ccdb69c6b3478e5e3640460c94218f --- /dev/null +++ b/transcoder_all/layer_3_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d248f2b735eaa345adf61f09e65d9d9e64dafb208f2d16262f08f7d2c6d603ce +size 84020088 diff --git a/transcoder_all/layer_4_width_16k_l0_big/params.safetensors b/transcoder_all/layer_4_width_16k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0bd6943e60de825f793a466cc99e418d1a196656 --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ceeafdfec3eb7189abd85e0215e0a6b5d3079e1465162c81e0932f44dab7a372 +size 84020088 diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/params.safetensors b/transcoder_all/layer_4_width_16k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..74f1c6e1274b1fa5629ee80f2d67bcd42f2b1820 --- /dev/null +++ b/transcoder_all/layer_4_width_16k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9bf25051407f249cfe9d571758c85bc4a90fca455020f77ab51cf7241d0dc746 +size 85658592 diff --git a/transcoder_all/layer_4_width_262k_l0_big/params.safetensors b/transcoder_all/layer_4_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..738f84d48eabbf1f92a9bf38e6af225c79bd1d2f --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3ad58d8b21810b9bfb1f27be17e40e86494cb21a40b82c9eac7633d4a1c84783 +size 1344277376 diff --git a/transcoder_all/layer_4_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_4_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dc8aa57a166ae2b1d90334daa7fe49c699713d3d --- /dev/null +++ b/transcoder_all/layer_4_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12731998e594d7de3e6d2af086a8bee8c1959fa08af0604bf7865ec284eaf598 +size 1345915880 diff --git a/transcoder_all/layer_5_width_16k_l0_big_affine/params.safetensors b/transcoder_all/layer_5_width_16k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ba3967ad18276139f87b9f905fe658ae3ab68354 --- /dev/null +++ b/transcoder_all/layer_5_width_16k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7b58479474b3d62aa8b363586f351ad6319e290533f8d715639f4ab1cf5eae5 +size 85658592 diff --git a/transcoder_all/layer_5_width_262k_l0_big/params.safetensors b/transcoder_all/layer_5_width_262k_l0_big/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cd1471fc433f69eaa2ab740e0d98bae2aa1c6763 --- /dev/null +++ b/transcoder_all/layer_5_width_262k_l0_big/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae08bd409ac705d3abe5ab88e8cf21bd996499dc9c9ff17533cc2cd38f9aff3a +size 1344277376 diff --git a/transcoder_all/layer_6_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_6_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..db4f6d4d30e48722c1454b262e9cb1288cb78011 --- /dev/null +++ b/transcoder_all/layer_6_width_262k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a29a351cd2813ea090af614c98d965770dd350a4d476c6993ca4734a23824798 +size 1345915880 diff --git a/transcoder_all/layer_7_width_16k_l0_small_affine/params.safetensors b/transcoder_all/layer_7_width_16k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..604348aa1e07c6c39e994ca4fb34c7ff4b755943 --- /dev/null +++ b/transcoder_all/layer_7_width_16k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e2d130327db32f993043ce217024e3a1f19b23ee12ae886da9df1903f8d5c9d +size 85658592 diff --git a/transcoder_all/layer_7_width_262k_l0_big_affine/params.safetensors b/transcoder_all/layer_7_width_262k_l0_big_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..384564d8637dd7cb4d8a0858d3c7996ba41cd946 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_big_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:571bca02314088aead9cc687e88a93b3334b70a7c0a9f2069b55ee6ed7dea1f8 +size 1345915880 diff --git a/transcoder_all/layer_7_width_262k_l0_small/params.safetensors b/transcoder_all/layer_7_width_262k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..534e638234cd7674220abc9595a86acf010f88d6 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3dc0ee02976bc0d4ac5d502368ea96cddf789a90045b666b3149430f7ef0269 +size 1344277376 diff --git a/transcoder_all/layer_7_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_7_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ab1a25efc784a23120c0609896f2a43377f03717 --- /dev/null +++ b/transcoder_all/layer_7_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb790a4208af1bc152f9d409bf07924787d7c5999661d9bf6b8266d6e33817fb +size 1345915880 diff --git a/transcoder_all/layer_9_width_16k_l0_small/params.safetensors b/transcoder_all/layer_9_width_16k_l0_small/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4aee1a73eed8665fe0cdf4f4a7986c83b7571cb5 --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_small/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0cf80fb4d09e938dd1aa7a1865588230cfe72c36060ef9e5d1b2a67b279d71d1 +size 84020088 diff --git a/transcoder_all/layer_9_width_16k_l0_small_affine/params.safetensors b/transcoder_all/layer_9_width_16k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c5555edc5154ac387593579c7cf0bec62046c38b --- /dev/null +++ b/transcoder_all/layer_9_width_16k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db95b754551ec32fe3d9e3bddd4e324b2b28a58d3be4a8b904664370a6d40c51 +size 85658592 diff --git a/transcoder_all/layer_9_width_262k_l0_small_affine/params.safetensors b/transcoder_all/layer_9_width_262k_l0_small_affine/params.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..09d2cd2a87e340b726d1c068de76826a6acfeee8 --- /dev/null +++ b/transcoder_all/layer_9_width_262k_l0_small_affine/params.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f1906468d88f639eff3062b92700cd777b0e3f2a792bd3ec9a6cc4e8c51639d +size 1345915880