diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7900fc56e6508c62074789aaabac9dc26d24df5e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/config.json @@ -0,0 +1 @@ +{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/lr_scheduler.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5441e5cff18f2f6f82f927744aebdaaf855c229 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afafb111606b64199b463308d054f28acf71d5cca5174e9fd9fce0b0ab0a717c +size 1076 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/model_config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b56d9a1cd00bd722f97ed9e4cbb48bde4e6d94f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs2_experiment_2.10", "dataset_name": "withmartian/cs2_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/optimizer.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..24e176ab443457ee52321a7baee198ea4240f804 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7da704ac6e2129487da820bdc9a9115f7090d98c6d111df2c5072620f9558e7 +size 255767204 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/state.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..267b29c9e723d3750c009a9951d454d236e713ba --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba2943dec258b4b72adf0815e0552de5fe634fb0d527143961d6ad61e507c078 +size 493312 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..eab697c6b29f7e2a8c2f0f10017a007d9c8d1251 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.2057291716337204, "fvu": 0.0022642838302999735} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5f7ef5d7cacd1a47a880efa06c537ee51d942f78 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70cba99086b79c82c1469918102f54834ea64a885e127b8c4f0e404dd31f31f8 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c92774ed39d1e6f1e2ab86a79b72e6031a6d795b --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0001302083401242271, "fvu": 0.01662026345729828} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f4d6bb0f0c9fe37e35a3b50226b3653d0be3a4c9 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a060d18c76f80717cbcadd1739bf2ebe3822223a37629b150192c21bd5ba0195 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c534022bb387d5e701a212688f96079258c3ea3b --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.01743653230369091, "dead_pct": 0.23743490874767303} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..534e102a1bf05f3fc38017415e9c18537ce64ba9 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b7c8c65b22ab6d392199b73e154ffd27b87758182da30543e4ed03a75bd09e4 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..57d51cdf28863ddc3ebb71e8ce59bd1d42de4abc --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0013020833721384406, "fvu": 0.027486111968755722} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3f2483f25140d8bebee13e64f468f6c1e31de63d --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f1bc41a20f300a09c3a6145546c775e564b409a1f115be8fdda8a125bf29543 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/config.json new file mode 100644 index 0000000000000000000000000000000000000000..efe5292367476b42c4dd1eb9546e5fbc2510a071 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/config.json @@ -0,0 +1 @@ +{"sae": {"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/lr_scheduler.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d6aa50a7aa478f20cb4ee6fdd0886d4b66d239d0 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a19ec7a7d99a9517726d577d84860d99635fa1571577af8b6d7ba1c019ae8379 +size 1076 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/model_config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b56d9a1cd00bd722f97ed9e4cbb48bde4e6d94f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs2_experiment_2.10", "dataset_name": "withmartian/cs2_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/optimizer.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..101984ead5b964299137a9b9a3d1f9685eea507a --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10de60f0daa1814700b5cc0c40e8fa2b906b2a0f4e9e0468a809fda0d05f1bf4 +size 68243108 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/state.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..71d17d8c1872181de1e493cb812b632f0f054d19 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d613321b12a736d53fc68af5317721a82e546fe2089aca98469f639ecab033f8 +size 132864 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..684bea357db88b67e98f54cb9874d7a6e186c2a7 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.660400390625, "fvu": 0.016979176551103592} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..eff7014dfd9a4bc30c524a7d97d5a2ea76425b1d --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43accc6b9f1c89519cc7024a64f0fc6ba6b4755c444cf698891dc289663cd2b2 +size 33575240 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..605705aae1421bafa85465bb94d214abf3642dad --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.2900390625, "fvu": 0.03907718509435654} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..53f3ee0c1dfbe6968b665921da34b79da181286d --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2c225481766b3491172ec156fc071c7b8993bb830073d79c57a445a8f41b22e +size 33575240 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fd87776e81a6d04588685f155e537747a991dee8 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.778564453125, "fvu": 0.05981138348579407} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f492cdc05ceb33d3c586fe13c53416c39be816ff --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:60fc257c06d4c8b6ce12e140d995c06032651e56630bfc79f430ca3b37a012dd +size 33575240 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ea559454c207d5ccf2551adc3de73cfb4acc0197 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.419921875, "fvu": 0.07065705209970474} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..eb64be13eee53a404216bd34c34b6ff26bb1f305 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d850eed5a84a0a25eec6c3f8e212319b5578021385b0c3ab9f815357bcb18ae +size 33575240 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cfbdfcbdac2b37ad623bae0bd3a6c5f992f3ff78 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/config.json @@ -0,0 +1 @@ +{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/lr_scheduler.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5441e5cff18f2f6f82f927744aebdaaf855c229 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:afafb111606b64199b463308d054f28acf71d5cca5174e9fd9fce0b0ab0a717c +size 1076 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/model_config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b56d9a1cd00bd722f97ed9e4cbb48bde4e6d94f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs2_experiment_2.10", "dataset_name": "withmartian/cs2_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/optimizer.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..9b0af1414927b828ced84a70fbbe57424417ff31 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd2cc81079cd17a50237e1762d19abe77309917d73e12122da7d44d550f2bbb9 +size 255767204 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/state.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..3bf5b4ed977d38a8dd1a623a76a304374602ba74 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4925e63d32f98c7d7fbcd1f289d91a58bb78c6f7e90016fc0cd87eab9145e36a +size 493312 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..af7b3e6ebbe9306bdb82a0db3a2865bd2bd62d0c --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.20664063096046448, "fvu": 0.0016410077223554254} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ea7dba5ce747080236095e2eaab552c7486a3ff3 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bcaa2f42fe77d37044b250f5569c5d855438badb960d0ed06b912d08b37e8c1 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..173321ffc4419a37a923b6bc08c95f9ac68240f8 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.012025310657918453, "dead_pct": 6.510417006211355e-05} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b97262539b0ed4ed7b69cb84a301d04aa6f9573f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcb44088b0363248a709436443525c29acfb309eac1b98970936a6085436c032 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1dc75c11cd54d89d854bc397701269341a9d62bc --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.009096871130168438, "dead_pct": 0.3643229305744171} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cee38daf20e376936f3c2f770d96c976e1852bcb --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9af7bb21b582bed761f367c258982d59d2f408b42736ed46c4b29c42456e805a +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c6d68570115508535c41505c13e2334db2440216 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.019307561218738556, "dead_pct": 6.510417006211355e-05} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ea20231f8ceb6e76ee6c551a238cfcea96afeaf6 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:891e8a78203bcfe42bab23479877cf9cb544e67001e8fefa5eec28795091a9cb +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3821fdc93d21e39a98bd0a96403a63f49aed893e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/config.json @@ -0,0 +1 @@ +{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/lr_scheduler.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..063c9b68a9ee368cab3ee017989901fd5a869ce9 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd74b3fabf7af5e0d248711a1fa05fe261368df7b69f594226dfd4b050dd08b9 +size 1076 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/model_config.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b56d9a1cd00bd722f97ed9e4cbb48bde4e6d94f --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs2_experiment_2.10", "dataset_name": "withmartian/cs2_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs2_experiment_2.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/optimizer.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b823f8d6c456386e5a105fac2989be5e84e83f7b --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b35ae95634215166143094c1dbb793f31646306114e2853746fd0a87a0a447b0 +size 255767204 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/state.pt b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..813b29f0a84bdf7eb08d0c042038a415d4ae574e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cb2422ce4383bbff63b4c951874b84a74af8d6d9dd8133875e779d868cdb869e +size 493312 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f0b5c9ddd9090d816042740b420236d888292470 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.7277995347976685, "fvu": 0.008592553436756134} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..23b1d65cea52d39c9e8f61269d059d38f36dab37 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0c7f2878d248d9a49a4c10287828721736fb71f7f07327e4cc71d72685c5355 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..03a9512a961c107e58470356cb436c0de096c18a --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.21985678374767303, "fvu": 0.022585762664675713} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..966984f97f15d4b6c5bf4590d4c3514ce0c04255 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea3d0a437f0da5be06e00fdf732e739e37c6742b6a832ce282dd5c759fadcd69 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c484bc27a39099edc0a5516d2a3229d25037e5ab --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.657421886920929, "fvu": 0.039229705929756165} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dbaaee4912785a628252cecef55799ee0f4711c8 --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e74b2b8e773e69bb54002007d321780783e2e4465b21b0b49080f1eefd1d324 +size 125894984 diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9a9775eac16cd8664506f04677173fd4d058d90c --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.23815105855464935, "fvu": 0.04541328176856041} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d3ca6637d8240f2cb3abeef40cb14cdcf9b49c1c --- /dev/null +++ b/saes_sql_interp_bm1_cs2_experiment_2.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:164b6a24c81e98f0ed93943d18c537740003f2535fc2a32fe2750b9904279370 +size 125894984