diff --git a/pythia1.4b/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_1/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_2/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_3/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_4/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11/trainer_5/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_0_step_9765/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_1_step_9765/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_2_step_9765/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_3_step_9765/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_4_step_9765/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_0/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_154/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_1544/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_19531/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_29296/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_48/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_488/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_4882/config.json b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/ae.pt b/pythia1.4b/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/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/pythia1.4b_sweep_topk_ctx128_0913/resid_post_layer_11_checkpoints/trainer_5_step_9765/config.json b/pythia1.4b/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/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