diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d93b1653309171e00e01771a698a07700acd22d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/lr_scheduler.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5441e5cff18f2f6f82f927744aebdaaf855c229 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128/model_config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc8c1acf1f3f2750a8e52aef5781f9bb2d91cd73 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs3_experiment_3.10", "dataset_name": "withmartian/cs3_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs3_experiment_3.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/optimizer.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..9e57288828f943ed534754b259a2790c003456f7 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05bc737c2f583a7284067f762d22f5dc411999b81dc3d208e7366c5aa0cfab12 +size 255767204 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/state.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..93a0d09bc6b0bb1e1ee06f51a44748bddff07214 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c556a21b6a0c8685b667ed3d31ea9263f623a5f9294cfa25ef9b7535f142138 +size 493312 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..62d530654fd6813cc63612c954baa927739e6e3d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.18613281846046448, "fvu": 0.0024563544429838657} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f9559da102887e89ab9e4b60f6b2c13621fa7536 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fd936b6c30c1ce6ea74d0cd9327e147cfc20375ed42db7c7234eed4794283ac +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f3c941cf01755448c221b980fe73a71cbaa2213f --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.016500595957040787, "dead_pct": 0.00032552084303461015} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c0835e19c2a5a64e2fbe6438c7953368a187f27b --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c304bced17a8af8c3b2ffc00fd849a1de346b67cbb6b0d76de57169d707435bc +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..623af158677c110a275960882e0acd6833aeae4d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.018058784306049347, "dead_pct": 0.20774739980697632} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f66889e7cebf3ec71c8892419d0a517553baf600 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:53833a29f9fbb82ea8e6ff8d993a9bb2ae98a713a068a1678a6c5b71fdb9f04c +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..7a5c49940485d2a23606145a39728bf54a47b7ae --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c454cbbedb309636eac16ad56c17fed6a2aa4b02 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0014973959187045693, "fvu": 0.026534151285886765} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1c09a28ef57a48875550e9ad832b108a9fafa9a7 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2a447e5988964f955ef0755905f556fd220dd6b8b93baf3b7d21407cb2ba0730 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c80fbba655a878ebd06bf5b357ac880e7bc04107 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/lr_scheduler.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d6aa50a7aa478f20cb4ee6fdd0886d4b66d239d0 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16/model_config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc8c1acf1f3f2750a8e52aef5781f9bb2d91cd73 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs3_experiment_3.10", "dataset_name": "withmartian/cs3_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs3_experiment_3.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/optimizer.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c936c93c7e0e59a5b539c45e6be06b468979f086 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d4d59e4af27494d7a55d71e3027cce99c6490154de521e57d113db32b90e221 +size 68243108 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/state.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..7e9c5ac95e45fdcdfa74a00c67e72d17bd814455 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3642f8e60a9d07d644099d3adcc60d2dedcb4afe7375d80861cda0cfd29ee212 +size 132864 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5ca00a95ce360f8af2ae2ac894ad5a66c95d024e --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.644287109375, "fvu": 0.017582185566425323} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0d5d236588a6211d2be5a6ac97886edc8d07c1dd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26ab021fd45966fe30335feff1fd87662d938d6b72a4aa94e1e931752961e750 +size 33575240 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9791d32ff701597bbbc4000956d131535c21eae8 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.233642578125, "fvu": 0.03958441689610481} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bebfa861bdd64a48429502711aee71a47f7d1c74 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07dcceee56acab72bf1a7a053e3f237485dd5adc1faa3b9e5344d425314f1e40 +size 33575240 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3801a2e134af6a857949f25d77fbfd1cf387ab7b --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.695556640625, "fvu": 0.0602182000875473} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..63cabb94c5fa5e8914ec1f8d2283da410484fa7d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d54af82c1e5d6ef611afc3f0e4625ac4da93868716073389f56b1946a8fecb6c +size 33575240 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8943661e77283935d0527f13a585ab60ac97bd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a1055d192ba5896e61f3d6b6477632e142275c63 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.3828125, "fvu": 0.07253600656986237} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bbf9805661668e80d67ee6f9aca0e2da55cce0cc --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc095dd6554a039e301b8a348a3d35c174e46616bd9461b855d56d890ff602a2 +size 33575240 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5eef6117ae2ae77d359714861435ff1a23b38cd4 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=256", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/lr_scheduler.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..063c9b68a9ee368cab3ee017989901fd5a869ce9 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/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_cs3_experiment_3.10_syn=True/k=256/model_config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc8c1acf1f3f2750a8e52aef5781f9bb2d91cd73 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs3_experiment_3.10", "dataset_name": "withmartian/cs3_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs3_experiment_3.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/optimizer.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b9f85c7965ccf8ed3897d0027afad30b738dcb01 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3af4b7f60df38a89e653cbb346214ec92efe3238f1f83c0506f30f2dcbe08bc1 +size 255767204 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/state.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..71c447702d933185ea8ad58e1f910cba8130e930 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:714abfbb80f4016e75d93e8f7ece32c6f752f64f5640a66e19e998afbd3dfbe9 +size 493312 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2fef8a7de29148e222e37441b7c185c58f87e9a0 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.22682292759418488, "fvu": 0.0019393906695768237} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c3f7afca818f45d960480f5495b23a8e99a8e84d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cf80c780554996ca1a41445ae88263fdd6a0367bf5f02e2b4bdd6c6339a56e4 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..09a86f85fcfd40626292cb3fa1683ebb05470044 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.011782245710492134, "dead_pct": 0.0002604166802484542} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..14838c06644480e57480a244d4be119dea71032a --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:32a79c0e4fe0116b797176be12f787629731e7fe0b3aca7f30b27f8528a1ff55 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d4119f4a9b66fe22ebcab567e1af06313ccb144d --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.008456540293991566, "dead_pct": 0.3470052182674408} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3416a9fdc4c6297ada2b905a9ad977c1a0e0cc13 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f27c8e1fb20b9b7f0a2f7beb9eac3fba227ef5328b9811d8331bf6f07524d222 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..ce6546ec1f37867d65385e3e774a35e3a73a064f --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..31cf9a76fbce88a12cbd3f46fd1171b2a2b0fb91 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.018147287890315056, "dead_pct": 0.0001302083401242271} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ff93e1841872ba5837cf9fac2d044f03c97aacf6 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=256/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a67514416df992b8eb31929388b991abb33a752f5f5171c6198a1002d6e041a8 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a7d57ea98505b65cce98f13ef484dddc6ca53cf --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=32", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/lr_scheduler.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..c5441e5cff18f2f6f82f927744aebdaaf855c229 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/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_cs3_experiment_3.10_syn=True/k=32/model_config.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc8c1acf1f3f2750a8e52aef5781f9bb2d91cd73 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm1_cs3_experiment_3.10", "dataset_name": "withmartian/cs3_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs3_experiment_3.10_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/optimizer.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..78441b1f1c534118bd916570106e0ac4bc6c3518 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff29307068acd83080bc5ebd9639a1e294d72b723dfa3060221025f8b02aca2d +size 255767204 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/state.pt b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..1717b2e9a91c6ad668444beae403e9520472bec1 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd6daaa1932aaef0e9cf0e8375bda643865024f1a08912b8539928cea4ade49f +size 493312 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..68b4b17cacdb975b843cff28ca889aaec43c8d9c --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.6492187976837158, "fvu": 0.006474225781857967} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..88762a01106c74ee6091c89231cb1df1c67b73ec --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:212d1387343b7211e42332fed218a1343655c62a597cdaae744fdf308febc7fc +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f66ce8cc89be5d9b9a1df7e6d435c3c550a3afce --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.19967448711395264, "fvu": 0.024956272915005684} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ee040fb19e7945c264dd7f0a0c99c1963db64abd --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.0.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1414458a67f329c4a7c62310dcdcce8f9d95e1c25b0fb260d06471e3f36a2687 +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7462646f6a9516941b319c181a74e99860ce1605 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.5365885496139526, "fvu": 0.04285687580704689} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d3bd0a45d8664c249f3dc5381dc7fd746c947841 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.attn/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b20f584bea288af164b28cfcc0dc31758b2ee734ba4b20bffd0fea1657918ffc +size 125894984 diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/cfg.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..d613f5a83c724bbbd58146cb4f50add56e3cea9e --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_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_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/metrics.json b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..eb976c5cc4e1c5d82fceae2af4eb01712bee9173 --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.04432789981365204, "dead_pct": 0.2031250149011612} \ No newline at end of file diff --git a/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a59d85c932938be30a19581aafa271ee7298f9ac --- /dev/null +++ b/saes_sql_interp_bm1_cs3_experiment_3.10_syn=True/k=32/transformer.h.1.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cc09b75f23eb2ead69025946798da038348a25be41ea0d2c27086c1cbd824e5d +size 125894984