diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..621a3edf08178ca7abcfdd8663e92f3b9687cae5 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f931bbd2be0cf27730cf6315612899804a86d40faa2a8acb8d6f4a2534fdaddc +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1815888b1aa43b5f00df3b11eb0dbca5137d9b27 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8992ebcffcc12c4f339eb1b2185f0e70774b6e5c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77c04025750f094a1207b5c13d9c0513bfb6738d7b56ccf82823b5cece85ec97 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..866b7d04651069cf78dc8b1480b38acd7a7546ea --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..aae5211f56aeaf8e0c04e3be36cfe839a84124a6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:777821e6b4ba99d790da876723dc8e934cb1723b8126a848baa763bb716050ec +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d12dada054272989fca8fc81c058c8646ce9d2a3 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d30c0b0d0532728acc5e95e5cf48de1bd41299e2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9165f71a2c18bc3a3c5db1008be308f7dff6bf8c89227924af9cb3065de69064 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/config.json new file mode 100644 index 0000000000000000000000000000000000000000..426927eec085801a63520cd919490cc5bb2545bf --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d7276378081deac1853e6c3970765d7939756b8 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd61e41012e58b8d9599f2cc17d064146bf530fae5c81420128b6c6c91dd05af +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9d1e0a8612711aaafec4d769cdfd90fcd051ba6b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..64a48a858fbc5a7b41a850c9368b977e9167dfb4 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e946949d76acfd70c4e05de0701fee087ab718e6503dcef0651efc0f732713d1 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/config.json new file mode 100644 index 0000000000000000000000000000000000000000..010178adc751bc114242bc51b7d30d271b7bafc9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1c1e0ea6cca2ce42042340b59377eddc0df85b78 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..622815b74ee93a9ce51b3ff8909defeb298a5b4f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b67e75ad65e356011f6408fa8110bc0781b2b342a7ded331c7dda307480c614 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..852b69e94abc523cdeed4e57951a565f63a436d6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..ad419f723165434c7774c9870933ad26ef3d1427 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1983dc5a21aef1fd81c3bc5f11f083c8e6eaa9702bce34ef9493c647b5884972 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..28d0c711b55e801f3a45a6146c6818763fc1983c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8f534ac7eabaef656fe6f8e55871fc0ae721ed93 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a6755014321c45b857963ffdf611282d3c2725576f5c0e106ad001843fdda12 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db76818cfa4a72905c9b03856b6893a0054b9700 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..c301e4239a9de714d9b04903e831b78ca083d55e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5845ca9123c5cd425cfb70f950e0d791bb99e986ed461b6d71fb5071f3d63dc3 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..af1dbeeb96ada486c1c2f36b543575015e2e26df --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e1d9bb084d716bb61952747aa5c9ba7817415d12 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07fe5d2d6477315d213d5eb4c4759a5186541e3be5e087512a23bb4f9cc12617 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a21b9c914eacbf205e07292853595c257aa2bc6b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..7f6bc0bebe498ef32a9f2ac4d04533110439c747 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1bec26256c8c3ba421bb3b42a864678b00b999101543869c1ba642835be4b156 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4ccad33007c16c175437b5ce46ebf6adef93b5cf --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..de343a8a27d6f364db0a4c6740fe6e990d2dfe40 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4b5711335410894032df7a45fdcc95e0c3288daae6dca00dfa501df123fcf82 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4735306dfd45298cbc0b09f8be16818a044ac009 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..700ec6884f888a3d5a9a9d4be1c3e53d2b67f353 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45541576a35a7e1167ea8acdb8d1ef865f24163bcffd0ee92c5ddb4a1527bd2f +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..00c106928bd8daeb1d1e4b93988885a107238c89 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9466a8489ac7c012a3fa51ee2d1c308c8910d013 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..9bae16923bb9e2198dc386ec07321ec9c448dbfb --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c4caef5ae1673ad808e2d7726be2ea3624c070025e17a6e3a70e988632b003a3 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2b96255028b0fb753e01f51b1b54497b4724c53e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..b6eea6118c5890496b93e08e07500cf8444d7684 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:baa3efb085b9a91077c06f6d7b69397242041feec0ede677f11f424c1fb30686 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7e42b955aa837e251645b36127f75f32a7b234c6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..c3279e277a617489e2e54a8c756e9c4d7861be2e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:990c2f403a04c9e4f15972f9f1e3d93eca79f878fd70d0a9f4c7571113040fa9 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ced60384242634385186657fd6c4b504231e34f7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..ee179a96c213d6b32de46876d38609d4efec00da --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcb7920bb0cb1115a3966b3a81ad046e9220eed41e9626792278914c4f8035a7 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7faf51b4924cf8e174492752175f0b5dc975ed52 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..f686fa6c1eb744d8392325c7941350bbd4c091cd --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e0c4fee3f2d78c9a5cb1c1016c3a54dc2dd72f88e3fc2ab09e63157679d8c07 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3bea0a0b2a584d814f6edb9d52e6c40dc403662b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8fccc38db6e749412d495c179c5f7408fa5dcd8f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8dd04b0baf20a9a97f5c63901f6d96b329fb9ca9f22167af7c23296fa087b08 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8ec055e5153cce259b19f487566ef13381c06ac1 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d418f849ceb8080fb966fa8ce07b331d1f60452 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9861440d2992b23f094bae25bb91a4d04513194387a2ad7ff333557396280ed2 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5721855be71afca662a1c301a65f98b482e10927 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..a101e4a8d4abf3a360813a6ffed171eb30ab0aed --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2270228e169e6210fe2544982668f8345b32d7d0dbe2c65127f650929905054c +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d6fba5467bd0a388b5cad6ff77efcd5bd7a5260 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..45540452b491d96e0a0a80376a2b529e4b761c2a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..5fa759ab10fbc88147147a5d5b3499f6088233de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:feba1d6daa2bd336e8b354e38a4282aac2bd2144f447e5a32e7c60e3cefd909a +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8d768942b7a0c20d27baf40ff08a0e8cfe673e48 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..1e46f574422fc4eef2ca3251642785334b81615b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:beff393395e9ad23b1d2aaf4de38b1558198083ede9adf716fd71631dbaf534e +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3c0fca409f461e520aecc70c8ba06d798282f174 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..a050957436e42f9ef4028610bbcc9ddd65472a04 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:115e1f639e6552ea27f3bc911fbcc62298f1b92f685d8ddc3254a94d0d406a5f +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..471665658c689495116468cc5f51909834021899 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e1acb3ae8102d18cf7ef8e5245260f65ac9aa01f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc6c20e2b5600e8c4c7429d9dd17521bdf325777e47058f73e892158fbd5004c +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e3d3ca32488b955a6448fd6df96a591f92c490a1 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..cb066147c65fcfb7ff7a086d40f7793e08b78637 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fb9b3e9e674c097d99e364f43bcca534d31d244cd398f2747f50b42bc10bc22 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d61b864748ac6c42496c80ffd1a125d5ef6ae5dc --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..0e715ee7cc7e46e122ec30f911698bb56cec828c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b5b3503c0f06180ca0856288fe896a9cc4893f155f5493132f2dec863462fb1 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..79dfb2c0be94c7fc8877e3a581c16bc6cc58dd2a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..21366aee2bcc30e26ea70585043513bfc5b4146e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ad60c35e080be1a2934db1dd764c9177ea9ea2755ae33a3660fa14f0050be7c +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a479153da515607b1cd73b0c5c54f2eef82be01 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..c732ab9bf97a28507d82be7fe2ae59eaefb45824 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf92ac2605db1e2ca8046ab4b6d8665df41ff9e82d5bb1b116249444243e77f3 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..52251623ac799e2d2345a5d4541efed605cc5d04 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5290130a6a9d5e5ef0d2cf041d565b31aa3e4e51 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..6b2b63829b8cbe8962dae5410489bb5270e25da5 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a06492382483c3b2325ff4c5de1b21bc54101b0cd67d0dd0ad25dac3b1d65f53 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..470cc04fc07a68265c3293e0e378778c494acd20 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..01dac5a7de1170a1c4e9d07eb827fe5dd550837b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:251c17c0c8a9d4d698939c2b1b4bcd883a35fbf9256cf79c54dfb37941833d90 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..00691422831d690ba2c707f95f09e03de9b13818 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..3c245b349a62b4891c30c51215819da51e65152e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3154630a9f9b5126e9e538a71020ba3b03465f7da8e0d84e14e62f0a4bddf125 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..871646d6cfc4720a70929934813da87ff510c7d2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..564f0f8989c7f674c004001d5c9ca460c0efdb35 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:170ff32609c6ebcf049ed8e4c9f9d0896f0d6953c728d308cc3ac61c16c0618c +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b060ca7d827d1d11437e23d2d78c59b54899b47d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e771481830513e25fbe71a031c906f4593a823f6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:183ec595ee48cedd7c55e24e61ca0ee38b1842dc748d9afecff84eba2b2b408a +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91c9064135851ff8f0c2cb7693b1dfbd266fd708 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d377ffd022ae15ee8b31d11372e5fdb609ab23af --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1974e4611ff0578b25846dfe6cbf88e1a71688d52622402456d7a8d690b8ec0e +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..03279a15276c93e989bac6cfe8b9af8d9dc75a62 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..80a5043c5caedc1effad93d8a48a693c461769e7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:46b03def18d835dd47f09461d940c425ac5eb7a35905139341ad6da97c3e8da9 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5c6d16f3a84b768d8ba5b67c6a6a6f0bc29fc986 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..cda8e8ef4a332049ca50636e27ee1d14bae1d1c4 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22cc86b0997e62c62357ca2c5e5af5f1819c2be6cd1aad7c4b637be46ac0ae49 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f2bdb1b85a6b2c575689fc7b2fa307f40bb8139c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d7b8ea8b27f909d326d8b2d650e26db1f0babb68 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ffcb84d0ec209752c9c7c1e3363eb9e7cbfc23f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:068ca08d4ba0d5506078032eb27fe5960c2c51ddeb2d14548a5e41e93bb4e4a7 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..301c99b7cbd35af4fcb030659befc2f90abddbc1 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..4fc02c4389c0a8aba30d5f4705e99d7917077df5 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ddf52fbf50c5effdb13951bb65fa6f1ab9e5b454b1a520948520aa31d64507d2 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..85a263ab73b294b10675417d5230cee3d1aa8513 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e44747a7dae8706715038ea8223795a911b832d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:952b705e510128c20fc9eadd8d641fc06e83bc5a393825d440b8dd7efad2463e +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..067845b078b1ba6f1089622eb000dcc03bbd1f6c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..5e5d427d3dbe320335cf797138e18ab4d8d39cb0 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:070094ff136d841a35505148e8946460315a753a19f7dd319d089c030c0791be +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7dd0bb5e8c54b93f8edd35e92e52510f18334be9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e9b48269af7833d43c011912d343471f508249db --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3418b256bea5f24c6029fec61ea885bac9b0c0baa318841e2521af117463faf +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2233b03eea218b36bf55757c04e1c9c9f6e4e411 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..0a5eaef81a760d47aa8b20c784897cdcb0fd644d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fad3dd8e69519ef9d87bf86629e17f26ac6a122268bee965ce924063c99d9bad +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bdb63d598565290427b9ecb6f2495dcecb593fd9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..0653dfde8d1137f2cbdf360e6cba703642576a0a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8617963aabb62d98414acb8dde8d3e41ee15d19f8a1701eb8f0796d58a58f31e +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..33a4eb587e5971cc276f3ac578ec3e9c7f80c268 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..db1f4f5b1214dcee83d55cda751b9bfbb7b2cf17 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2fd8c09ef803123fe8d25cc5d118176516444d1a8a3f044da607398cefe69f79 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8bb1f4549ee0db67ebd25aa8a0192c1ac2d44ea6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..295a4770318c69432baf9833d6f132e28d3edfbd --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..cb5f55c25a12f4af08b728b28ced9d6fcf2e11f0 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:362680264f8a22bfe18c3f5f89a2523b3f0a01b7814f34387dc1841a8838980e +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fde281317796effe28f0098e9fe7d766239bf60a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..fd63693d54636d186b4eefa58b0aa3fefcdec874 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8e6cefec95e5f4d7745ff828916921d748fe5bf2837419adfa47dd3a9031149 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f77f06e9d314c448f7dfa3c4e31bcda301e20fc4 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..2eb8b7912f63290ecd5e1dc1d93c2bc8f4ac6119 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e33e41b3763169bec77abce1d04d59463ff74cb73cf2b1f5b9314b62626f5ee +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3a75304f9c5656054ee8b31e9cc70412c1c52a56 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..80f24b574c365ac6f73c0f4b75d11aa3d9f15166 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be30cee6a2bde83ac9f483f1c8dddbf5a1ccc4ac9e0ddcd04eeb72f0d26a6b8e +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5964b0d9451c8db691f6cd822899a57cccd915f2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..a0ebcf513d0662f0ec8d47c3e29b9fd16ffbc86a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:009643a776613d8c6b41814f88cec9e8c9d9002fbd5aaa034c4019c4446136a1 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..85c2251ee77f6337637e66080dfed317a0e45ad7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..69c9126162cc0bbd1e29398ee8c1f6639b834b5e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a63646b530d37de15c6e3587c53e72c53d7817412073d3c6c807f84052fc390 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a556fbeb383ee4ffa0fff4746614544b2e6c89c0 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..32b1dd3f9848ee60ec34aa1aa54423c1feecc3d3 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50b648940f3cb977b237c11e845a7ad0d8e8d6c285d5b6d6b235c7ae872cb35c +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7f720e26c98aee7a3513462913ef82007643868b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..0ab04d4b8cb1ccd197f2da9b2c4ff38e845536e9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd79dc28ced687a4cd89e4bf264fe99dafdb1d7aeb9b98b0b55ba4193c9a2979 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2c8309eb58f926f6b623573c26e448efe46794c7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 11, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_11", + "submodule_name": "resid_post_layer_11" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..743d4afb7750bce777f07b7ae6cf1fce33796584 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d8cb489649db783348c82f6faa14cbde54500c3ec7941d99d7234273c564f21 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..70864e7fee1544c55e64085d38de47f96cee18ef --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e2c6370cfbd0903e3348b2095cc069d09677a74d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:020d38b7cdf4ac792c45bf3a28fd31fbfcdbc70a19ba416552060babec64a6a8 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2f025a9d58b7afd87db81341411047ea5e62c0dd --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_1/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..90782d5cf303870f2c5d650393398ac095159e3f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ee10864d7ddb90be834d2118151c34c70c5de9d1739b030b93577b00de8ef4d +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d64c9be300252cb08277bebf93a8a4c7c91204d9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_2/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..932392f569ec605091bc344b3312d16322778553 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f734fe6813e548bdfb96a5fbc786c059f05bb7facf62f60b9f65c0458874ae5 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/config.json new file mode 100644 index 0000000000000000000000000000000000000000..db3ecc74f334fb7785d5bf31f2fb6af8f6df0f6c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_3/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..106db6ec5bef65ddebfd617ccc399cc2e45fe890 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:557637e8841cd3bc5ade7fc4e7ce8b33dd5d18971e4ffc3261452b873dc3d542 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/config.json new file mode 100644 index 0000000000000000000000000000000000000000..21afc9ee784ad399bbe2c46835e7bbcd0bd4d9d7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_4/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..7d3f276a974fa6ce561452d8effce125069a4650 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dbcbbd1d9b86f303f62369830ac994444808d8b905f17b0c36934d42cdaed160 +size 268511016 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a71c324cd32935139dcfe012683e60df6d6ecd7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19/trainer_5/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": 48828, + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f0dfadb140cdae83dba866a70b42d66f3da7644e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..9f5dd56f440165be775aba8a7acd675d682ad7c8 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49ae6cc7c2e5090ded2794060f985f240e3e01730e4e221bf9cba225a0e762b6 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07fab6e32ba5592deea371265967f287e7d6ba0b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..a48b2086973cdb7c6010a313f96e1130aac3b405 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2b39d5c76ce30a11252d324fc606f4d42d9e44c54f131f208d508b8382a2827d +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2ed1cc9eb0ad297b065d65e199d8635f659d7eab --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e2c9b20a8060736cd2e96d5a140df33b72b3eecd --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0faf1717f9dc149c29655af7ae1d516dabb6e8ecf01bd8f27582a0761a381920 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..95976f0c6c8456deed45c1e0abcb9c6e8ff6ab17 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..b30bdfdaa07484e76b7658b3ab735e977adaee77 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e49cd8d58107abf58f1a2a26832206931b8879b533fcaa0c2612bf31818354b3 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f44a517aa6b548cd9c39ece8b6df077cc63f1325 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..b26e6b46332f5124683b16e611ceb705d4175730 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:86c08566b67d218533bf4957d9924a29b886ace0466507a14c4533f6295ccbf8 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a24aa47a76d3d1b107e227a4d27ba63c973c480 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..5cbe81aa468ecd8c06b3bfdce7c51e4808eafa21 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e88f5b96308dc21b9c12f977aabdfdd36a94a22ff4beeb209b79c297fd8b32fd +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5af2d783f0cc2864276dfab652dd831f0e257435 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..3765ec919933f15a4f4b5415aa38c2cf45cd5226 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24fcc1a435423bcb887ebca8ab36933b47c38050790da0ac1468fca3f550d11b +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e03e73e7be0b3353f00d7ec7574e58f74de15877 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e305c2bf07b3464c42e8ecd8f33d5f3ce9f2f29e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd9a778a163a784009c2078182a2ea20a06418ba6b826f19e05b5f39b6e3a417 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d69dbebe764fec7d558d7fbc061e89820a71284d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_0_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 20, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1938959390e4a1c08dc1cc1cbe765f65aa3b6292 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..b2796b2e8ff6ac6d717b3faf601c4b49a6109df1 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0820aa6dc6f63d91af066c5b492eefbb1afb73e40db25516aa98fbd52171c002 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..32edf4cc292295ca8106b7cac1a047632e936576 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..730c34f6f7f84d4de4da40cf1b03d440225941ee --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f90faaa4673cd71d2ca51a73eeb2bda71a5e7bac68bc08878b5b76df145841a +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2b5d2c17a47b879991beaebac9e2d38f4017ce93 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..9929f3d2950375aaa8fb9f3bca8fbe93e279728c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a228d3f3141b59d2cd519cf15020789d7f8f75d5b5d2b348ed9e78338c1ecbb +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..401d54c136ce52cb1263110b246639954ef5f87b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..db6114f62e7b208d77d63aad59e1052aefcd0bde --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c02c8a16c2234086fb37823a9de78b39f36f233f271634650f7706fa70c2168d +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7370eb386bf6ae069aaed9ee1d03e371d47adcc8 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..1160a5afba55473f516f2cfe0a64e74e17b38c7b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03476b586e5846cac630b8e9449c71042e5cdb6995350cc6089cd9a156822416 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..121e29aee0b8d6e148377bb10648848e870ce746 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..5a3cb4ced1fb8951bf3d29e0068dfe3f2eb50dea --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07a18d086d40d94dd72df12820a38166e1e36ceef0a7c440a0fda8c7550149d6 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fce8fd3a8eb016ad2677e9a109227043e72abe0d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e68af23e49fe2403a5d5d7b32bfb6cea7c3d4ae9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e555287c5dea9dc1ab7aa03e4ece43674674740f5a04786da34c6c9a2a42787e +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c8424eadf0d66d24d2397b587ae6b2d0f6d11e03 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..5fb7d2cc63a8bb90761fdb16a0803a64cdaf9305 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea23dce6b66da2a02c60aa248807d0b18d98f2916da0727b2a96e0467f773bfd +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..18530846e6ea2b3f1085b51a62c49c6ae524408e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_1_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 40, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..76a1a4cf0d8799666355f0d2151fb0f9245e56db --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..c21f60c1ce1d1b43c7dd69dfa3e917ff4f8038a2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50015f0cf41aa6f0437483b54548dac8f4095c2c5d22519b1beb75017cd4049b +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cff5ed1d9b25b59d8b8f99c3b55f219d3c3ea19d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..1e90fee853ac9f2607642a4f6b5cac7707139f51 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:540d013f5916d7671823e9a0e42e2a046c0314a18b0ca241030c9df5ca6e4a83 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6fe5640ffc64a372fd2f8618cf4ea6f31486afd4 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8035699efa1a5cac62846318fad862eb6a5a844e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3184bf48e175955e08fdbd4560dab1c773a7024fd5ca79532135ed3a573c471e +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..68108ded666046f5220e4d5606255ea7cb0f8aaa --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..15ce7bda33a9cf58df745975214a1d1b6cbe095f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4e3851dde8dcbc0890743139faabb6a1d5985a2bbf2cdcbe4384c918cd9c037 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d06095dd804da025cb1ef6187d6da84f608242dc --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..1e0275a723545e0ed243f904b23736cd6848601f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:064a9cfcd466d7e436942305233866b798640f5f8b0ee67a026cb0c08a5593b1 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d850e66e2c4fb2c75e594ca7ab184faf8232cabe --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..c1387988316e7fd70a72d839b80aea9c4f2344e2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e660cfcba9fa9ac07c0b6d20d47823b37d988454661b830bbe79cc07778e263 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..343a42bd6b7db5dcb8feafc3cafdf3c638c6c77a --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e20ae1b4fcb0de4aa59e1393ccdcea2c4e5cf33b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45c7c53936e634fed8ccb6b5cbd58ababc7ebeae1c4ed0c884b54a35c4f7232f +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3da6f418b87d831d6f39ee51acb89ecc38a39663 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..62b0cc1e3c0c45511d16c0cf661fd77bd86fdf1c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8765b9ad46735433d2bc9127468b4cf09c33472a47d2143ec4652473efa8ba86 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e386845b160619cf95492c52b140bcbd7c79c558 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_2_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 80, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0e17df58ad4f14f21fb529865a3a30c99261ebc6 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8eea805f1f1e7f11fc2dd864db258414a7b1af82 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c015376b13395d29ff439f91d087bd6b84efd12c1b90ff89e68704e16dc64de +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c9c44c799c94e6302a120d42eb62af41490e5b5c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..43ef115a729ed2605ec54c9b9f9d4e6c41a1c828 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1a081b2bc508ad6d06cd5795ce93157fc83bdf777b6d4f73e123faebe760d718 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d444641d4641a3f4684a00beefaa3fb641622beb --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..30a0507f5243f4c3b309a3feac23beb0c292f67b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb7dab7d340c6d40d4907b2e7d63df573a35ac1530350f4bd52c9ba64b8ee935 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7150c3c9d9903e29646f1719fe4b6567698341cd --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..212becc539496e948f71fdfba34f0602ee45f76c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b6e0b91c70d3b489b4f60ce7c2fccd14ff3b0d6c10819a4ace0b4481439e711e +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fa7c176d1a361861704a60be35a31b59bcc87a07 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..2da32ed24f754b139ca7f054467f777a5aa36844 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e40e2936eabbacb7d0ee1400ca593dd3e32ff18df182bed8f8b3de0b90b1e51 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..61f6a13ab070550269adec277f3d0897fe2b9759 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e726978f7a3227a4b11e224e6117300d8a6168d2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:518ed92560b123ec483a4bad641f7c412fb7642d21f3b83c86da3071e3ebe900 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..061427bb4b7f84e3d786b47ef01608c84f23ae25 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..7f94b3ac3bdf3d747fb33e35b05a685d9b8858fa --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5db024a395dbf31af0b1f51b7c2f3814e09e7bc79101ceafe0dedec2763cdc89 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..360a02e03c2f5c89302459e588b58e5633a8d8a2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..8dd778cfda3dfce1d6aa5c0e98ae919ed49b748f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c587086fadfe03409b64eecc95af889119202fedec598487e0c1be8a75eecfaa +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..08d5ae4e6b435dfe776b2c5d12547e780265f71c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_3_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 160, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..48e8084cafd6776fb8d83cd27b17340f3bc4c3d9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..2d999de1a5f5584ac718ff0a6ce99de28275763e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:403f63337d101110dcada4d7ebfc8574337c45d0f2994f97f5410d6d6d638aa2 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..99f81de6fc60d320c80aa553a22fe02fc2eeb8c9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..e8bf166e3b2dbaf00d82abef1a7b6f8f4291e853 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe9be10dca39715e71c4ea324d46f7e243636708b70f9059b801603f047a45b5 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a5d691ca2da89d7a22293dbccf768223bb7bd145 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..44742e9a439ca9fac55d9d776bd0e2d9e3655f1e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9c11b8afc20a055ee9416d2b8c6e9493abb4580db15cbb33091d7b35da61461 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5b16cb4a1ca534bc79214aea38d9548706723e1d --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..056f9a6902ecb16c2f39363d73949c5a06b385fa --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:597112002bd70cd282f873ecf391b2dbef9c2aa9d4862b5f6b1385e80624096c +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bd2fd11f4dc7c208a75b64964af5f2ef4dc84766 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..0b2cac584c0f5ec51c01120d0efaff2eba50ab87 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:331657c5bfe145affb88e51e9cf88b8d4c2756de54cf66153248ea19f8be9796 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..41f4ee8324aabfcd0b34e2a9e66be47913517f7e --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..f691b86fc7c2ec0c71d1b758c22c957efe52f742 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e06c45f41f467ee165bfe4765303cc929e630fac485272dc50af03696870e023 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6c3b795edde08fd903fcfb350256f0ec82e0c921 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..00b763d8aa3517d7e184416559384b2ea72e1f02 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b9e5167a684b21adcdba54ce31f1e2ae900df82fa69c4d6b5cc4a67a6db5d6f +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6f083d77c8ff5c2640b8ec774b04f55ea99d1246 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..f5e1ec677f169dce123d6d70899b987751dce464 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59a8ff7f7f78c505f908840191e0114251ac1b210bb8acc8c83aa5e7e7796d8f +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8b56dedaab88ffa2fe3905be5ef9cf6e4c5a7734 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_4_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 320, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d925e8ce01233786f37010f048bc4313242850de --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6adbecf89ab611c06dd7362aeed0c6223349b95ca84867cd6c4fa1aaee62e51 +size 268511032 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/config.json new file mode 100644 index 0000000000000000000000000000000000000000..770ed59fc66841fc7ce7b635177dbcdd79baa789 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_0/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "0", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..44cc1482e8eb8ca21ccfce7c17be22f89f7579f9 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e09b36fa53b6a6460bc6371a7b6f8cf6d3e4e5d09580e4a057e0189ed725ed63 +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/config.json new file mode 100644 index 0000000000000000000000000000000000000000..51be1547bab649a7d2930a61f49b5d3c424fe32f --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_154/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "154", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..6351e5d182278f024ed8d9663acbcf208bf68f2b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32b90d12e22fcc446971a2eec778f783cc70fc842d89409e212ad3a76ae17633 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f4eefbb5ca97b19341509a26e08af8e33489b129 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_1544/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "1544", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..4557a042fe728fdf247c7cb8e743e4a5353738f4 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5d8f94810c6b5f76c39b5a1272d13b82ee3582b58560a10acc2a0b4a07dd276 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4b35451d4d267f7deb4239628913ca873b660165 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_19531/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "19531", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..73d322df4d8942074c918ea68bff87a4c3d8dbf2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:df0cb07a1b33a43205b40f75d89384444c3b2814cddf7c746f6ae916e8626023 +size 268511320 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5e1321c7e4ae5dd580c0a2cf376a898a78234a20 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_29296/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "29296", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..833d25346b33f1f6913af37f9f8df9879c859c15 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9eafbb19009c06a6186a2ea645f1bd55981a1eb17fc3ea5f1244848997a98a1 +size 268511040 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/config.json new file mode 100644 index 0000000000000000000000000000000000000000..315d9bad6448c304bf3437537745f7961874c50b --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_48/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "48", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..d1788b263c461e810d17a07cf4119a59de79c275 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2cca8fa4d73a3a16a6ed14c2c019fbf852047b1a84c45acedf9b490b7349f2e +size 268511048 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/config.json new file mode 100644 index 0000000000000000000000000000000000000000..270df037112cb27e78fcddd15e629bdc435db830 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_488/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "488", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..22ba079f86498d6f7aa2aeebfee63737367d8cf2 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac69413f9879203a2fe635dfeb76da8781fad35cf5191af708e45ce1ccd2a9ce +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f268c586cb2ada54f8b36bb2f4dae20265fd42f7 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_4882/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "4882", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/ae.pt b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/ae.pt new file mode 100644 index 0000000000000000000000000000000000000000..4723617ff5606d89576fa5c86a33ebd82a714237 --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/ae.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:512deb9fd0018f9cb24d94cdc5f4ff6a307d5dba023990e1b3a55c54cf679a45 +size 268511120 diff --git a/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/config.json b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6e7961ae459242567c83a6692b51653b2afc631c --- /dev/null +++ b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_19_checkpoints/trainer_5_step_9765/config.json @@ -0,0 +1,26 @@ +{ + "trainer": { + "trainer_class": "TrainerTopK", + "dict_class": "AutoEncoderTopK", + "lr": 0.0002, + "steps": "9765", + "seed": 0, + "activation_dim": 2048, + "dict_size": 16384, + "k": 640, + "device": "cuda:0", + "layer": 19, + "lm_name": "EleutherAI/pythia-1.4b-deduped", + "wandb_name": "TopKTrainer-EleutherAI/pythia-1.4b-deduped-resid_post_layer_19", + "submodule_name": "resid_post_layer_19" + }, + "buffer": { + "d_submodule": 2048, + "io": "out", + "n_ctxs": 2000, + "ctx_len": 128, + "refresh_batch_size": 24, + "out_batch_size": 4096, + "device": "cuda:0" + } +} \ No newline at end of file