diff --git a/transcoder_all/layer_0_width_16k_l0_big/config.json b/transcoder_all/layer_0_width_16k_l0_big/config.json index 5f05c887f2cd322ab9f7971cbdd8f0834647cf2b..5e51f74e97bec05a015447b8e7ba7c8fe3efdccc 100644 --- a/transcoder_all/layer_0_width_16k_l0_big/config.json +++ b/transcoder_all/layer_0_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_0_width_16k_l0_small/config.json b/transcoder_all/layer_0_width_16k_l0_small/config.json index 414a476adc45f67b1bfb9834fe1b4d10447ba9ac..035f5108af93707bf834002182802c497a75d037 100644 --- a/transcoder_all/layer_0_width_16k_l0_small/config.json +++ b/transcoder_all/layer_0_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 10, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_big/config.json b/transcoder_all/layer_0_width_262k_l0_big/config.json index 26286312dcff900e314d0d622bbc0dc7bdae5f51..7f125ff3ffeb29c15401bd466010d5dca48ab239 100644 --- a/transcoder_all/layer_0_width_262k_l0_big/config.json +++ b/transcoder_all/layer_0_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 60, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_small/config.json b/transcoder_all/layer_0_width_262k_l0_small/config.json index c68474c7e77b4b5434702b85f401c8bd47dde5fe..6da2176349ced46076f2f54f2beb7d14d393920c 100644 --- a/transcoder_all/layer_0_width_262k_l0_small/config.json +++ b/transcoder_all/layer_0_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 10, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_0_width_262k_l0_small_affine/config.json b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json index 46ede8bc9b101392e80a98cc3db8c8b25e650041..eabefe6ca7c97f8339f1083a5fdd6447a69b37c9 100644 --- a/transcoder_all/layer_0_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_0_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.0.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.0.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 10, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small/config.json b/transcoder_all/layer_10_width_16k_l0_small/config.json index c7bc7d16b2e1bad8ba71953d0405398c58646954..80c7c5995f10006b4d28cc88b730385a45802a55 100644 --- a/transcoder_all/layer_10_width_16k_l0_small/config.json +++ b/transcoder_all/layer_10_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_16k_l0_small_affine/config.json b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json index 0bf901c6e802a655baaccc20b7b7b070403fd8a4..cde4969be78b43c31ec4af8cce75f911506a58fd 100644 --- a/transcoder_all/layer_10_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_10_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_big/config.json b/transcoder_all/layer_10_width_262k_l0_big/config.json index 4a74b39be808451ce4c7971b2926ca12ccaef7fd..63e768a38af3d03a8188d7965370c7a337fc38bd 100644 --- a/transcoder_all/layer_10_width_262k_l0_big/config.json +++ b/transcoder_all/layer_10_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_10_width_262k_l0_big_affine/config.json b/transcoder_all/layer_10_width_262k_l0_big_affine/config.json index 372a03d3d5d0761c3c2baf4a8eb2141d82c9bbd1..e97795245da69cfa283222c555ff0717c9b24a5c 100644 --- a/transcoder_all/layer_10_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_10_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.10.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.10.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big/config.json b/transcoder_all/layer_11_width_16k_l0_big/config.json index 740f06084d2c3e7105a3fd39368af3c7628f7f3c..f0505fd35ff223c05fefb2f697b81f24d1574551 100644 --- a/transcoder_all/layer_11_width_16k_l0_big/config.json +++ b/transcoder_all/layer_11_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json index e9bcca5d710a1b729a177e5db4c0ce154da9793e..7eaf0be76fedb4da0a4e24d0d2645ccb55565fca 100644 --- a/transcoder_all/layer_11_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_11_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_16k_l0_small/config.json b/transcoder_all/layer_11_width_16k_l0_small/config.json index d6664c1449a2243b7f221e3993b863af1b8f6ead..b7c5fea1ec7fdbef51de185d23c7360fe0eadf13 100644 --- a/transcoder_all/layer_11_width_16k_l0_small/config.json +++ b/transcoder_all/layer_11_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_11_width_262k_l0_small_affine/config.json b/transcoder_all/layer_11_width_262k_l0_small_affine/config.json index a4eefbc969ac1f4660a5b1363967362504953721..d3fa4644ecc5d6cb090d07f73e429a9d6a41b983 100644 --- a/transcoder_all/layer_11_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_11_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.11.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.11.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_big_affine/config.json b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json index 846878442e727e5c264c563caeb869916415887a..831567349703a85092cb4240884485c0ee921076 100644 --- a/transcoder_all/layer_12_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_12_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_12_width_16k_l0_small_affine/config.json b/transcoder_all/layer_12_width_16k_l0_small_affine/config.json index 3672cd330403a2c84d868c59ef7c347394018161..67b86e46faa3d05f27106d023cc4748b7c7d3bc1 100644 --- a/transcoder_all/layer_12_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_12_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_big/config.json b/transcoder_all/layer_13_width_16k_l0_big/config.json index b84c94848af009fddbf3ab0db727b280e9927a0a..721777780e9c7c360c1a62806004aaa0d99ba7c3 100644 --- a/transcoder_all/layer_13_width_16k_l0_big/config.json +++ b/transcoder_all/layer_13_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_big_affine/config.json b/transcoder_all/layer_13_width_16k_l0_big_affine/config.json index afdc798549b97009efac37e0370adaf51a0ed165..c9acbd10c43661f146dd5e4f8b3aa66ed52fea4c 100644 --- a/transcoder_all/layer_13_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_13_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_small/config.json b/transcoder_all/layer_13_width_16k_l0_small/config.json index 262492dc91ef2592d061110a6231c5236c5149b5..7aa8bcdd3d8400ce08234e81995c326056b5431c 100644 --- a/transcoder_all/layer_13_width_16k_l0_small/config.json +++ b/transcoder_all/layer_13_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_16k_l0_small_affine/config.json b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json index 8cb31b6e8587f65a901ec688c683a07792b3d6e6..973779e72dec04360c6c672ee2b511262fed30c7 100644 --- a/transcoder_all/layer_13_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_13_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_big_affine/config.json b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json index b5ffbda30dd4a53de90742d5cd812b5cd91b3e27..c3c282d2f2b80df8403e91a718a3ec68919a245c 100644 --- a/transcoder_all/layer_13_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_13_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_13_width_262k_l0_small_affine/config.json b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json index 45497ac6fca2df91ebfcf9357d030c4c0ef9130e..7dc21319b6213b524c035f177f1cf9549d78ca48 100644 --- a/transcoder_all/layer_13_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_13_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.13.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.13.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big/config.json b/transcoder_all/layer_14_width_16k_l0_big/config.json index f0299f1b668985d78f4dd6e519bcad65f5fd5784..ab322db1bcf65e64c92ef86fb0a7a00dc088b16a 100644 --- a/transcoder_all/layer_14_width_16k_l0_big/config.json +++ b/transcoder_all/layer_14_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_16k_l0_big_affine/config.json b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json index fb4a629d2d5dab001fb096737a8e0dd4a2cd59ea..30c1d13fd61af42f9d099f1c51c1623eef112237 100644 --- a/transcoder_all/layer_14_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_14_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_14_width_262k_l0_small/config.json b/transcoder_all/layer_14_width_262k_l0_small/config.json index 0e28238caa46d4a4236a16fc239bc623650e8549..cd5b847745ec55d23da03cbcb69b425368d9cb7f 100644 --- a/transcoder_all/layer_14_width_262k_l0_small/config.json +++ b/transcoder_all/layer_14_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.14.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.14.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big/config.json b/transcoder_all/layer_15_width_16k_l0_big/config.json index 65eb5e8b62d2d27d6de7ef5673fb29a72b8b988b..05836d041119cde904a0abde44fded71cbce756f 100644 --- a/transcoder_all/layer_15_width_16k_l0_big/config.json +++ b/transcoder_all/layer_15_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_15_width_16k_l0_big_affine/config.json b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json index cbce2b0629a441485e1adf32f8e8b7fd943a67f0..69ddd562e5e130871079008ea31ad421278e9ce2 100644 --- a/transcoder_all/layer_15_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_15_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_15_width_262k_l0_big/config.json b/transcoder_all/layer_15_width_262k_l0_big/config.json index 4d4e8bd735f433072c7824f5a54ab3164dca6600..6d1c88f7a4a51cfff1081f7476c6e1dd611b6b87 100644 --- a/transcoder_all/layer_15_width_262k_l0_big/config.json +++ b/transcoder_all/layer_15_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json index 7c5021c6bb7ff035adb0c1c8c87c97fffca4384d..2b0b3a862f4e7b55021ba3686ad3b6c07b1cf988 100644 --- a/transcoder_all/layer_16_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_16_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small/config.json b/transcoder_all/layer_16_width_262k_l0_small/config.json index 9fd14d6057802b7aa67ca3a47be8333957f0fa62..d92eabf1c6a1f920495272cb75f03015c7aba9da 100644 --- a/transcoder_all/layer_16_width_262k_l0_small/config.json +++ b/transcoder_all/layer_16_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_16_width_262k_l0_small_affine/config.json b/transcoder_all/layer_16_width_262k_l0_small_affine/config.json index 66765a97681ecd0519c28b057078e3cd01b6faea..7420b86bc81f3e6bf3e885ba3a4ee792b948db1d 100644 --- a/transcoder_all/layer_16_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_16_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.16.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.16.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_big_affine/config.json b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json index 177cc57e5db98f8c29a46c15008133e1902ec95d..5ebd2e583b6ac47cbf1d2447d21a32568da20e99 100644 --- a/transcoder_all/layer_17_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_17_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_16k_l0_small/config.json b/transcoder_all/layer_17_width_16k_l0_small/config.json index f84364b118ec6a5f7a24241c374bce101ade4d0b..e29e85123c3df1580fa3a0a0e73ff4118087bb01 100644 --- a/transcoder_all/layer_17_width_16k_l0_small/config.json +++ b/transcoder_all/layer_17_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_big/config.json b/transcoder_all/layer_17_width_262k_l0_big/config.json index d79c14a908fcdaff93ee5283b41ae58850ae71e5..9cbbb8ba094891475753deb147f15c0d9a428209 100644 --- a/transcoder_all/layer_17_width_262k_l0_big/config.json +++ b/transcoder_all/layer_17_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_big_affine/config.json b/transcoder_all/layer_17_width_262k_l0_big_affine/config.json index 49afa2066a3c6109ecc5f470d5628b68b1b8fac1..42f4b6251be7e75d1f85d2c96d41100d2dc2be65 100644 --- a/transcoder_all/layer_17_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_17_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_small/config.json b/transcoder_all/layer_17_width_262k_l0_small/config.json index 2bdf83eb0affcdce740eb3c4a93fd95c555496c5..da625efa2143ede5021df785c62114c96a6ca81a 100644 --- a/transcoder_all/layer_17_width_262k_l0_small/config.json +++ b/transcoder_all/layer_17_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_17_width_262k_l0_small_affine/config.json b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json index 87837f0f48683ce76205b0247f8bad2c078e0a09..fa1dc6801271d5d2c76c78317956b515d5728ece 100644 --- a/transcoder_all/layer_17_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_17_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.17.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.17.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_1_width_262k_l0_small/config.json b/transcoder_all/layer_1_width_262k_l0_small/config.json index 761d8b09f26c15212ee2b878eec1984bad0c2f1f..38d73652eeb0d31ff3766c37b856937a215e5d26 100644 --- a/transcoder_all/layer_1_width_262k_l0_small/config.json +++ b/transcoder_all/layer_1_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.1.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.1.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 11, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_big/config.json b/transcoder_all/layer_2_width_16k_l0_big/config.json index 59956fb39e7e760d6ca28ccbef5baf420e212d1b..99214cdae48ffa63174c98161147e2c3197ce543 100644 --- a/transcoder_all/layer_2_width_16k_l0_big/config.json +++ b/transcoder_all/layer_2_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 80, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_small/config.json b/transcoder_all/layer_2_width_16k_l0_small/config.json index 7c7b63b4814c90a1b5c27d722fe3fcd3bbf913ff..eba569652bd7ce00248e2a91c6d8994cf221e780 100644 --- a/transcoder_all/layer_2_width_16k_l0_small/config.json +++ b/transcoder_all/layer_2_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 13, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_16k_l0_small_affine/config.json b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json index 4d4a6692996b62c874d36cfa10030943fa852d1c..e2f6c676a24240c24026bd70dd1d8bb1dde05dc0 100644 --- a/transcoder_all/layer_2_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_2_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 13, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_small/config.json b/transcoder_all/layer_2_width_262k_l0_small/config.json index cc685ed3b38b88e5918529659d97c1989cdb2fba..c40a2403c71d99480c4602c1dada71497b85b15a 100644 --- a/transcoder_all/layer_2_width_262k_l0_small/config.json +++ b/transcoder_all/layer_2_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 13, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_2_width_262k_l0_small_affine/config.json b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json index 928f7596e2ebd9ef32ffe6b76a1a82518edbb01b..281235027670cbcb8ecbe05d9c1afcc58a333a3d 100644 --- a/transcoder_all/layer_2_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_2_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.2.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.2.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 13, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_big/config.json b/transcoder_all/layer_3_width_16k_l0_big/config.json index d8868a4537b6d6b73f15a0058402a99c7b2d9e1d..0086d5849c631922824ce0bb6e41048c1ed58028 100644 --- a/transcoder_all/layer_3_width_16k_l0_big/config.json +++ b/transcoder_all/layer_3_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 90, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_small/config.json b/transcoder_all/layer_3_width_16k_l0_small/config.json index e9168d3296823870ec33acce9d896c12eb9a1a88..643081c994defbc633a08d3db641fe5f0895ec0b 100644 --- a/transcoder_all/layer_3_width_16k_l0_small/config.json +++ b/transcoder_all/layer_3_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 15, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_16k_l0_small_affine/config.json b/transcoder_all/layer_3_width_16k_l0_small_affine/config.json index 1d3e1b111129732f7f94f8d6cf531637579b7c9c..24cf091fa3336d0621d5ef92dfb9fb031d22496a 100644 --- a/transcoder_all/layer_3_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_3_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 15, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_big_affine/config.json b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json index 178c2199dcce5173b3384d02ee01dbb194e7a6b7..f2b915e2229ee72480ae07f6f9cf9c8de1030d15 100644 --- a/transcoder_all/layer_3_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_3_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 90, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small/config.json b/transcoder_all/layer_3_width_262k_l0_small/config.json index 15a2efe44f9cebc1f2a40d7ed4d3744755c06bd6..5b9045bcb803187a50c63c8fc7a5e0283c9ce22c 100644 --- a/transcoder_all/layer_3_width_262k_l0_small/config.json +++ b/transcoder_all/layer_3_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 15, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_3_width_262k_l0_small_affine/config.json b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json index b0c40b7cd180c82d9b2f8571898a817c11c6ab3b..c1041eb464ffabab9fb2d33ec923e6ba105442b5 100644 --- a/transcoder_all/layer_3_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_3_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.3.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.3.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 15, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_big/config.json b/transcoder_all/layer_4_width_16k_l0_big/config.json index fc37d81bc02489aac384926fe7070921e233a401..a669b090d7bb4b6c9a5e2c26f94507b63717fd42 100644 --- a/transcoder_all/layer_4_width_16k_l0_big/config.json +++ b/transcoder_all/layer_4_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 100, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small/config.json b/transcoder_all/layer_4_width_16k_l0_small/config.json index b4960dc9f27c0cac88311012beff4429e258aa4e..f6440bf933dd4a156273c32cc8e5d7b452152bd8 100644 --- a/transcoder_all/layer_4_width_16k_l0_small/config.json +++ b/transcoder_all/layer_4_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 16, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json index 905cf5130e2caac580c6b11ae9d52cbeeb5b0658..269ca00ebb72642953f7987946920eaa9e1d2147 100644 --- a/transcoder_all/layer_4_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_4_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 16, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_4_width_262k_l0_big/config.json b/transcoder_all/layer_4_width_262k_l0_big/config.json index d8451898d0648d2cbf923445c38b6526fe7166b9..52630f41bc8751b8ca176b53ee4486f74a09dfcd 100644 --- a/transcoder_all/layer_4_width_262k_l0_big/config.json +++ b/transcoder_all/layer_4_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.4.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.4.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 100, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_big/config.json b/transcoder_all/layer_5_width_16k_l0_big/config.json index bdc10be7c5d2c578132710ca179e9923a9bda022..8f14f16401e10fa731bea6c3ff78010ca3d1e1db 100644 --- a/transcoder_all/layer_5_width_16k_l0_big/config.json +++ b/transcoder_all/layer_5_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 110, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_big_affine/config.json b/transcoder_all/layer_5_width_16k_l0_big_affine/config.json index 0c2429ba52427cad7a07a1849dbf296a5b1c1463..5bfd0fcb406367cfbfbeaba23eeb19dfcac21fd1 100644 --- a/transcoder_all/layer_5_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_5_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 110, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small/config.json b/transcoder_all/layer_5_width_16k_l0_small/config.json index 4ab893bffbefdd5c03150513515e55d75b6728b2..8caf1abc868b42bc506e3cd16b73185abb4a6ee8 100644 --- a/transcoder_all/layer_5_width_16k_l0_small/config.json +++ b/transcoder_all/layer_5_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_16k_l0_small_affine/config.json b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json index 3b652a2e502f76c3ba30ca603075ceee996ee02d..a04e65893c64dd388e92f140ba873d016059c125 100644 --- a/transcoder_all/layer_5_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_5_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_big/config.json b/transcoder_all/layer_5_width_262k_l0_big/config.json index 4e622e46244623db9fdd63a1ebd455ddeb8c5052..55f340025233de0b4a705fc266b19b35e8a6b0b1 100644 --- a/transcoder_all/layer_5_width_262k_l0_big/config.json +++ b/transcoder_all/layer_5_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 110, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_5_width_262k_l0_small_affine/config.json b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json index d45f1ab35b0c4af730fe3a444749450def194010..f0d1fe25ab531ef5773dce810cce5a596c9cb1f8 100644 --- a/transcoder_all/layer_5_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_5_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 18, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_big_affine/config.json b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json index 86fb00ca7cdf393eef51f33bd284e0241b3cb4dd..c2e0898043665bcffe0e8e5c1ea73f38d3642c32 100644 --- a/transcoder_all/layer_6_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_6_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_16k_l0_small/config.json b/transcoder_all/layer_6_width_16k_l0_small/config.json index b1f4fcc0e9be5ec87046b3b97158589596cc3c6b..94e39ed2a813d9a34f8206501402d762792443cf 100644 --- a/transcoder_all/layer_6_width_16k_l0_small/config.json +++ b/transcoder_all/layer_6_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_6_width_262k_l0_big_affine/config.json b/transcoder_all/layer_6_width_262k_l0_big_affine/config.json index cdbaa533680b6014ba9dc914dd45e337da9f6367..3fda431faba2ecfbf9d6c4ae7dcaf9316f297a64 100644 --- a/transcoder_all/layer_6_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_6_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.6.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.6.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big/config.json b/transcoder_all/layer_7_width_16k_l0_big/config.json index a08478eb539d04688f9db561784526ac434f4af2..5cda1e60f858990b285479673c76526ff85ebaf3 100644 --- a/transcoder_all/layer_7_width_16k_l0_big/config.json +++ b/transcoder_all/layer_7_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json index 75dd140bddac3cc7ba7f9667de5d5ff42236c16b..f036f3db14b35de5155b20445bc6fd2d497d314f 100644 --- a/transcoder_all/layer_7_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_7_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_16k_l0_small/config.json b/transcoder_all/layer_7_width_16k_l0_small/config.json index f3147c60e29a2db14d3206d57a5e4f4433ae5521..b9ba4240331d25cfdc8f83d0ed5747f7fcb23ead 100644 --- a/transcoder_all/layer_7_width_16k_l0_small/config.json +++ b/transcoder_all/layer_7_width_16k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big/config.json b/transcoder_all/layer_7_width_262k_l0_big/config.json index c06890469d04586a80461801a50d1ef035117fbc..81abeb3c7f41ea68736d4d17ae4b4cb9446f3692 100644 --- a/transcoder_all/layer_7_width_262k_l0_big/config.json +++ b/transcoder_all/layer_7_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_big_affine/config.json b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json index c51c4d1c6c433782c894b1372356a1cabd185a3f..b8a56d8fc7e1b02734f9baca40e14475574e50cf 100644 --- a/transcoder_all/layer_7_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_7_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small/config.json b/transcoder_all/layer_7_width_262k_l0_small/config.json index f2a9c1d433a0d3393e281d25a1a0d46a15dcc1d7..6eec4a54d7a905c77a7a2d0d11df6e2ded346410 100644 --- a/transcoder_all/layer_7_width_262k_l0_small/config.json +++ b/transcoder_all/layer_7_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_7_width_262k_l0_small_affine/config.json b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json index eecaf91e52a76f02ea72e75adac430ab77e9b849..5816a9f26c0c185e02e2034a9bab7f101713dec6 100644 --- a/transcoder_all/layer_7_width_262k_l0_small_affine/config.json +++ b/transcoder_all/layer_7_width_262k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.7.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.7.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_big/config.json b/transcoder_all/layer_8_width_16k_l0_big/config.json index 867d6d93c55ba5cc20bf85dd4b05ccaec57b71f6..e117d6425a1ea9bc9b8f9e77e7cc957acac1fb26 100644 --- a/transcoder_all/layer_8_width_16k_l0_big/config.json +++ b/transcoder_all/layer_8_width_16k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_16k_l0_small_affine/config.json b/transcoder_all/layer_8_width_16k_l0_small_affine/config.json index 5e52cb2557c08f6e2cd413b20ca1e6ad0a9c8cbe..192dcc40e1a4d4bf011d504d8d5a0487e3d6473f 100644 --- a/transcoder_all/layer_8_width_16k_l0_small_affine/config.json +++ b/transcoder_all/layer_8_width_16k_l0_small_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 16384, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big/config.json b/transcoder_all/layer_8_width_262k_l0_big/config.json index a33e4af8a60f0a0af7f5fca6d45c2171e9e938e9..eb9417c69fdca37d4f0ab93afcf53a31df3a007e 100644 --- a/transcoder_all/layer_8_width_262k_l0_big/config.json +++ b/transcoder_all/layer_8_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_big_affine/config.json b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json index 7932a819cfae01ef73d20d6a9e615a0f7ddc158b..83c34064dfdb3ff2ba3ca26557a42489c230773e 100644 --- a/transcoder_all/layer_8_width_262k_l0_big_affine/config.json +++ b/transcoder_all/layer_8_width_262k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_8_width_262k_l0_small/config.json b/transcoder_all/layer_8_width_262k_l0_small/config.json index 530b3e795e5fe9c1d95cac49777f2b154e5118fd..b985fa1a32ff8ae49c755dca4426d6b0cb43ed13 100644 --- a/transcoder_all/layer_8_width_262k_l0_small/config.json +++ b/transcoder_all/layer_8_width_262k_l0_small/config.json @@ -2,8 +2,9 @@ "hf_hook_point_in": "model.layers.8.pre_feedforward_layernorm.output", "hf_hook_point_out": "model.layers.8.post_feedforward_layernorm.output", "width": 262144, - "model_name": "gemma-v3-270m-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 20, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_16k_l0_big_affine/config.json b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json index d021c494a62bb4203e8f53e1074318fcfc3d307e..7344d2211c7e61bdc9e57ffed7d2c17c0ac1de6a 100644 --- a/transcoder_all/layer_9_width_16k_l0_big_affine/config.json +++ b/transcoder_all/layer_9_width_16k_l0_big_affine/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": true + "affine_connection": true, + "type": "transcoder" } \ No newline at end of file diff --git a/transcoder_all/layer_9_width_262k_l0_big/config.json b/transcoder_all/layer_9_width_262k_l0_big/config.json index 0b4b87dc239e8e8be82515cc6aaa7f170d478fc9..63b2357fa941964a79061c784b9e74c9c26235dc 100644 --- a/transcoder_all/layer_9_width_262k_l0_big/config.json +++ b/transcoder_all/layer_9_width_262k_l0_big/config.json @@ -2,8 +2,9 @@ "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-pt", + "model_name": "google/gemma-3-270m-pt", "architecture": "jump_relu", "l0": 120, - "affine_connection": false + "affine_connection": false, + "type": "transcoder" } \ No newline at end of file