diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e602868a7ffb3871ad277c552b85965b53306e0f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_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": ["model.layers.5.mlp", "model.layers.5.post_attention_layernorm", "model.layers.6.mlp", "model.layers.6.post_attention_layernorm", "model.layers.7.mlp", "model.layers.7.post_attention_layernorm", "model.layers.8.mlp", "model.layers.8.post_attention_layernorm", "model.layers.9.mlp", "model.layers.9.post_attention_layernorm", "model.layers.10.mlp", "model.layers.10.post_attention_layernorm", "model.layers.11.mlp", "model.layers.11.post_attention_layernorm", "model.layers.12.mlp", "model.layers.12.post_attention_layernorm", "model.layers.13.mlp", "model.layers.13.post_attention_layernorm", "model.layers.14.mlp", "model.layers.14.post_attention_layernorm", "model.layers.15.mlp", "model.layers.15.post_attention_layernorm", "model.layers.16.mlp", "model.layers.16.post_attention_layernorm", "model.layers.17.mlp", "model.layers.17.post_attention_layernorm", "model.layers.18.mlp", "model.layers.18.post_attention_layernorm", "model.layers.19.mlp", "model.layers.19.post_attention_layernorm"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/lr_scheduler.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1619aeb0b307c4dbe1e57034770cbd69eb75c48b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c1a4ccded57e94e581be291e5bc725b8c8e883138bcad563998f929ce54ece4 +size 1652 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..60fae66868c7d709830ace1320203cf0cc0bda52 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.011143377050757408, "dead_pct": 0.00758928619325161} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..201a966cc0bc2a13b129ea5fbab4dfdfe1fd20b5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26b70e6030838b0a5139b6d3ad6b6ce3a919ee94dd20ecbca513028fa96adfcc +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..718a93c11f2d817935595b3bc14287e2884d4672 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.004017857369035482, "fvu": 0.008762086741626263} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..385db68c8aeb2818409ebd433c3c9b89c75793e1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.10.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d87920aa769096bb31889dcd100d7760890cb9c65ba7fa49ed4bf74e6ebe798f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c2a513f56eff6642cfe0b421f0cb787ae107cbd1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.012227058410644531, "dead_pct": 0.0096726194024086} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..16068bfe82d94790d5bc287b4a4f66725a92630e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:00530f95e7f09543bedf83233ad4adeba0d0adf97a619d8ebed51aa1128077da +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..bfe421648d5a157a6287aa01aaed01e46a198728 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00513392873108387, "fvu": 0.010888863354921341} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b687ea05fc99bdc896af19eb1c3cf0de091f3498 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.11.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7e6a4919d9b0aeb08d3314734ab3ec46a8978780fd60c0dab53c1acaa1b7749 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b75d6bf8ab598166533cce87a2e79c99a24443e2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.013765087351202965, "dead_pct": 0.007663690950721502} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c4612b6c4333b6f7d73945b8752b32c5cbd75f67 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e16c225c2d47b5a6615914d8a4e62c7d863373a0fb078c0cc7ff0b1c6d22e20f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..14f38495331fe72aa4ae9e3f53e860ad9741089d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.010856058448553085, "dead_pct": 0.01004464365541935} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c3883bdccc426bbe5cd204dc9a9ef7c5a8100b7f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.12.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b266fcccbc2a12baeb5e9c14f7434f4c725e4eee7a7c5a23df51d6a6e256f7b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..067424b36940246fe18164a3cba621f957b9d909 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.015192780643701553, "dead_pct": 0.0055059525184333324} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..721f19ebf150971e7ca5e605bcb1b0407c2ee3e8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe702ea2a7d8e08337c980bb9eee72d9149c8911166f237721e364b18d250035 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..32e5f48fe09fcfd982084c15c07aa0e5683802f0 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.01202225312590599, "dead_pct": 0.01004464365541935} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..59e05a540bcc5cad39a19b2fc1b4d90bcb762ff7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.13.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c9c78a5bf72a2681a3c85b450b37d11287976c167a19536b8207101ef6105bd +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a68a9d2b1c1752f63b37a2676d8e1e07f27a4bcb --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.01987934298813343, "dead_pct": 0.0012648809934034944} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c6ae82c5575b6b63f01f33e95e32615853d7118d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6e3f2d1a8870029e1f3a890637a34c7f65336744156a07a48b7a6c1de1c9117 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2b297543ea3b3ded3f2c87d65217db23cf97c396 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00907738134264946, "fvu": 0.014239244163036346} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..671c28448daeced1e750bfc8fa5ca243a8077171 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.14.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c71198823012d8575bd5aa4ab8d92ccd37da5b9b265a06498777e0b5624ce0ce +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4ef83d60d3fbfb66c8d31dd9d2c4ff4098a54d9d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0005208333604969084, "fvu": 0.023401465266942978} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..085849b56d4dfd7d5ce853614e8c44eb8eba0e84 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:082673861e33fabb4d467840ba2e8598256763dfc348f4647e2760a37cddebbe +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..21146e11096bb50c56101252df16376dac619281 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.014060208573937416, "dead_pct": 0.004092262126505375} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..957bb632e1ccef2a7a05e38cb356831a512a7ffd --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.15.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:924e63b29c8a439afdd9f18892fbfe0899b1e0d110307cf54a6043b60afadf41 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..50ce5f2a49af81800edc7adaa5865e6d8c7e94cb --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.016603531315922737, "dead_pct": 0.00022321430151350796} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ae1edb582e60eabb81933fc498584eb2c075f0e8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14b234d5e11026e3beafd482b1ae7b14571996abc94e5d2b72f92ea83cc74308 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a09f3610f55ec91f1fddf1456a18c61d72293adb --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.013535319827497005, "dead_pct": 0.0027529762592166662} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5cdcfe002efa0fe5f97c1c7e468c203415537450 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.16.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bdd4138f7126c21b3c2c52502c86a18c7b14d5ce75cf66d5a73d2e75b63e7656 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d329672a874e600da2caa779719f9916cadc1ad5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0014136905083432794, "fvu": 0.011707901023328304} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b3217fa00a2766b035bdecddb0a3ec0c81a29196 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d31d0403c99aeacbb92a2b74e829a0edf56793ff0136d6c6f7f4ace994546c8 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..af206c19a7ffa6a7b9c4f92512df70ac1ba5c36e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.011492160148918629, "dead_pct": 0.002306547714397311} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7c5d0c11d9d84ad7e3778002e4661044324a5ce8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.17.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:526f3402b3c6843b76fe8e83f1159fe894f13254d252872ff8ae0be67c7ac82b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ed252d7c518d1c5f687e53529c0ee1bd44dca06d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0, "fvu": 0.015863902866840363} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..11930d32b6b25474fc610d319459f24fa1dbc9d6 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5f0d3424871b23a14367deddadf769138093b8a7648ee901858f3adf5311a67 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8d7e6f7a5af2692e5d864769e52fac3c33904c11 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0006696428754366934, "fvu": 0.012910332530736923} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..412ca3c5f20ee35f16816cb2d1827ebf8e755415 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.18.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8d71c249d55d4d0a963a70a88d3f567c1da64a7941ef39e6d53098024e447a2 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..49a939119a919940e345e7c0a939085033d24e64 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00014880952949170023, "fvu": 0.015174311585724354} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9066824347dcde625f712b71ed63532dec51c0d8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e306994a36b48445298deefee7782c59af463e613b0996bc62a77b5fce83f2c +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6b86ac04be85246f02d8d84d97e6f4f54e2e0830 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.011404115706682205, "dead_pct": 7.440476474585012e-05} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ec3e70686b088c410ba91202112562b83c8f2e71 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.19.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0af0d62ad16102b412e41fb06c4339b58531062aa0ed364e029d3209cc6f186 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3e80563caadbf05465795c3d3d8f1d7e973c4cee --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.004711732268333435, "dead_pct": 0.006473214365541935} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..affed3f4db3d562bbf276f718ace00557ee49db6 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac68bb6d08d1487e7e063259dde4a65134a8dc1a68c54d6211cb205c9b7cf52b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8ac6d24954b8a4b20a3783a29ab26ce4547f2b3c --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.005395248997956514, "dead_pct": 0.0016369047807529569} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..66790f388c1f2dba516fdde5cda71a10b91cfd73 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.5.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:500119928adbd83ff45f02fcb588512ff28094036d8f32b8f94432188ea53577 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3413e5365a254fe851cece95e0da8eda802a0443 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0020833334419876337, "fvu": 0.008038644678890705} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..39eaf1f5c6bd0eaa4836dd28cf5e63ab17f4073b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7fd5b2cce386597957af3ebb6f1ddf26950a6174022d3535674f13a7ff120e7a +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fa8be551a09935422fffd95511165d7d1c615b00 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.002008928684517741, "fvu": 0.005864235106855631} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..775b6e7e45b20098a77da55514415277f91d7716 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.6.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a41496fc546d36e80b0e506d3ad6f356f56f64278ff51421c282aed98b12775 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..d9d020075b44270539241d4cb2a04aaf795bf1e2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.008542937226593494, "dead_pct": 0.002455357229337096} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b710b476b3e50006b6436a2c0336c910a41ec30b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d16148199142a6c0fdbd3e9474ea6be1587c43ae6a2a08bb8fb31a474baa2237 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ac24a1cbf020878d6d09f674ee94ce2425b88a57 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.005462926346808672, "dead_pct": 0.0009672619635239244} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d0a34a7308e0800496708350afc886bece065001 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.7.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d65af97d49b20a545c511a64549524a6e86286eb37799e95023f5204e2771cf +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a38c120e6cb5d5624915423ede87dd3fb9b34c33 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.008527319878339767, "dead_pct": 0.0048363097012043} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..87bcc0926fdb4cc0103326fca9e560f05dbe99f3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d6ceb1a264b3b6822b51ddd5de84d4ec3359c7d22241c0bd8d17b8e487157092 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..beddb13b0cf9177b96978a55e51e6bb60d730ee2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0025297619868069887, "fvu": 0.005849564913660288} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..696d12ff3cae0df88402f7cb820ce4cf26a2a7be --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.8.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1e24c674738b7644bc469e46f31b9ca1039b161d5bcb77cb5ba27394eebca72 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c5f7d43fb66b2dc96df1aecf4492f4e7d8b6cbf6 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0031994048040360212, "fvu": 0.010036525316536427} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b99b2d6d77d275474afc348bf447b5000db8e63f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1a97bf21a31f90a51e34caf9d333560ac6c2cd2b749fea3ac5b138e897ba9be +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..8655981effb80369bfa02540354ed1ea53920cf3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..7877c590022d232c6fb6643af7531f3305fc84b9 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0034226190764456987, "fvu": 0.007193866651505232} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9a29ce5047ae0ab6932d0306e6a1194d683223e0 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model.layers.9.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd81271e973a4ebcffacde5d9517e3f05c30568531ff4cf44437c2fbb12fcdcd +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model_config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a96bcf4f590864a5b571e20ff6dbb25da603f0e4 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm2_cs1_experiment_4.3", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/optimizer.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..9f02d5db54354bf1c36f6fd327ca1ecaf769606f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:010c497943d41e43b3ddc2a3e77ecdcca93c43c3116867f5317d1f460c6aeafa +size 1468809018 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/state.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..f864d3a295c1411f4392132d5ecb21a314b734ed --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=128/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5bebb4e2775c1a8899a141974257e76f388280211e3f04ca2424442695d52f3 +size 3233272 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/config.json new file mode 100644 index 0000000000000000000000000000000000000000..395b6495bb8945c0298bdc77c7365079a09dccf5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_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": ["model.layers.5.mlp", "model.layers.5.post_attention_layernorm", "model.layers.6.mlp", "model.layers.6.post_attention_layernorm", "model.layers.7.mlp", "model.layers.7.post_attention_layernorm", "model.layers.8.mlp", "model.layers.8.post_attention_layernorm", "model.layers.9.mlp", "model.layers.9.post_attention_layernorm", "model.layers.10.mlp", "model.layers.10.post_attention_layernorm", "model.layers.11.mlp", "model.layers.11.post_attention_layernorm", "model.layers.12.mlp", "model.layers.12.post_attention_layernorm", "model.layers.13.mlp", "model.layers.13.post_attention_layernorm", "model.layers.14.mlp", "model.layers.14.post_attention_layernorm", "model.layers.15.mlp", "model.layers.15.post_attention_layernorm", "model.layers.16.mlp", "model.layers.16.post_attention_layernorm", "model.layers.17.mlp", "model.layers.17.post_attention_layernorm", "model.layers.18.mlp", "model.layers.18.post_attention_layernorm", "model.layers.19.mlp", "model.layers.19.post_attention_layernorm"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/lr_scheduler.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..4dceeaf3df0b35c80daf7324b1896244d7b14f52 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c342c3270b5cc3fc57fef361246a67b03ad4c70ca9b2820108ee95d3b45d568e +size 1652 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0641464e01d2f7bd814c9950abf92667feb9ba64 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.007389636244624853, "dead_pct": 0.005059523973613977} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..889f9333ec490494005e696b875f0bbb287c1b70 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b40a68aea0d43a106de8eb009b3238e731de4447307d08cb99a5612889df16a +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fafce64beef9fb5b45ccc17cad9a051c25874070 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.005201994441449642, "dead_pct": 0.0014880952658131719} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..859289475491146fd8f1596d5307e560aab6fb78 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.10.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87c7957f5eb7fb9bf98b6b5491f893a593f0795ade55dcb1adb17dd67bb62209 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2a5a74f44a51718a72d9e450a0f91f8f81150574 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0078125, "fvu": 0.006918993778526783} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ef1a144c3fb7801354c837bf4c66d476636c82cd --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:078831e2ba1df26fe7585e8ab4cd3f6f3b36bc060251ae4177f4175cc60668af +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6b53226fa79606977aa1316c59f874c69c3d93d8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.0065234145149588585, "dead_pct": 0.002455357229337096} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f2ec7168b051d18a5dfd4d7a8b47c2118a99eff8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.11.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7fcc08843344cffd4a45e7bae6923864fb06fd08c1393ff9c05aaae246fa6ca6 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5bb7f57931878ca608dbfeb725b81610572ee20b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.007560656871646643, "dead_pct": 0.0061011905781924725} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..59a61a286b66acc84b6860768cc3366a025638dc --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1940c1e132d6fb947a2c090cc2f8b792aa20220c0893b506881d61c76634686f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6d5b38a8de589b6922647779fe2086fa8bcd2fc3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.006123366300016642, "dead_pct": 0.00364583358168602} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..57d5908e974c6209ac2ba1f6bbfb440aff5ac310 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.12.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4e7fd095bc5b2afd4e9eb18f8b2f75aff0f8590e0ce8534a5be205943b9f25b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6dcabb12b8b68ce86494e72528da7050ea4a5707 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.010443536564707756, "dead_pct": 0.0017113095382228494} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..67cd904ddc886e2f93dc12f662fe09a6b284af5f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:116dedf23d2970f9e9352820efc24ffd207a5160fabdbaa81867c9d07133686f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fd49e2ab3c91e93dbf1e111d48dfcf15956af963 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.002455357229337096, "fvu": 0.009360969997942448} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..731132fb296dd9c064bf732211a4c38f024d2939 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.13.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10c4fef82e368e044e1ad1f0b5aa9e681c6c9683d917335b383110e6a8632587 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..720f87c5d954793dba426c76842e90328f97cb03 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00029761905898340046, "fvu": 0.011759531684219837} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..046a0154dc36ed0f5b4c285b2c48cdf046e45c65 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b4fb6bfe120c679b6710c672813e4970812a896b8c8b87e9590a94e8bb1ad8b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8a5a5c5ebf879e3324d8458a48479eb128c5b6b5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0011160714784637094, "fvu": 0.007400806527584791} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4368bb346679c9e6dc915bf6915328965845a69c --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.14.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e575cca3db438e56a8f282836b9971fdbb1a791e569f210138d09f0edb4b019c +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e851da25ab43b6d9bb501c37b21ee98022cb67e2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.01267997082322836, "dead_pct": 0.00022321430151350796} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bbfb119a2d0b0e35bb5958dfb666cff859568c49 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:724c0492f1e5b06fad02c23e3f3caafa884928ac23a58d7d1f41ca9be1aa05f1 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0196b413d0dce982f902e0b26601dc8fbaa25992 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.009873402304947376, "dead_pct": 0.0006696428754366934} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..83f95d8dc60b2cca478cde2a4cf072420e1122c7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.15.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a85223332475f11a530304b752d959985941b079abffc47483bf1612f781c57e +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..eac74d6b266ddc02fd8a3dcb3b6a04ce9a81c317 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0, "fvu": 0.012106050737202168} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8660a658e6485d3d866c322180e9a095951a7a71 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ec3cfc6942a2a5fa5d90e4f32c8925dc9a71fcf137639ebdb4be193e942d381 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..6085913f8739ee1a77bfef7d9ad15a1276ab01bb --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00029761905898340046, "fvu": 0.0077447108924388885} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2461b83de091881e2f96c881052e3d1aef633f5e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.16.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03a9840841a0f35ef786166ebbf6a4d89da9ff6d9e48b911cfd178c58ec98304 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..295f6addbb5fab4f54a773c17cf31e8b655e2de8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.008571564219892025, "dead_pct": 0.0} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..65830e15dda05a575f09ee277a6eacb49bb40429 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c8a6fa5b58d9a748a0368ca8a25d684e5a16dd2d9e74e720609fe86c6cc7337 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..a535c5c5437f95d9b55e029c2d5efea81287fdf9 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0, "fvu": 0.0074108862318098545} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..570398433913aed23687e1d9f47a092a05819a01 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.17.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0eebeb967091378334b35a0318ce37cee27ab7342cc97ea1ea660b2d4829b011 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..05c7ce0abca2bde9e708e24316d4dbb2f509bf0a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.011035350151360035, "dead_pct": 0.0} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..60f451b37aeabfb721f388822ca6e437618946fc --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c333da59ee407b0b008ba34728a8ee1f532d43ec2970e0bf729476eabe8c68e +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..016fe9fbf34bc786f18cd8ad58b6985e992cecb5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 7.440476474585012e-05, "fvu": 0.008274384774267673} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..097cf8a440f3d1eb2a606411232b57af3386a286 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.18.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:09a63c32d7171afec417249398033ce24966b76e4460e1cfdf82915c34f649cd +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8d7d1785d043d61d9fd5d2cb98a6af4029374c54 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0, "fvu": 0.010861103422939777} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1f0ba5326815f2106acd4419b44d7c845ef2ed5e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa0fe9081df8cc5931a732a0eb5e7c539b8e9dcfd3591d567f5fdb1751ee95cf +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1691ac020c33f65c1a2133fe8dab848ec67d19dc --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.007869679480791092, "dead_pct": 0.0} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6f23a8f11b5c112a7ffdbd5d3499ce0e79e90844 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.19.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d099b4efdac6e7392151e142524de6ab1e1db491dc3eb9e7b002930a230738dd +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..3fa351aa68a9631134da6577d4a495d49ea614e1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.004021646920591593, "dead_pct": 0.002455357229337096} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0640a520e2a1b1f134db8bf283a5e04fddb105ff --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aebc5acfad1ece6a69ee687b0b5e13bdce84bb309a9fc1cf8e2a0a4dea63874b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..10c10f4253c8ef2ad3308f49cbba625bc6a716ef --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.0038359190803021193, "dead_pct": 7.440476474585012e-05} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2250f2469a4d434ad3fd7f68f4390cc2be0fde48 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.5.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e076203e63e199fe849be47df58c7bec5bff8dbdb0f6a05c83645d7cd55adbe7 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..2c986b6aedb4d3c88acae187404ba74f992dcdd4 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.005225897766649723, "dead_pct": 0.0009672619635239244} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bab4ef1a5b568ecc7d5fdacb00bf97330f59ec14 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5734e1c20f8f27927399ee0a178975fae907f4747f2447648fec9f1ac60f4183 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..4381e8c6baa4868f7d56ddeaf7323d626c671ca4 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.00029761905898340046, "fvu": 0.0040362062864005566} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..03a900dd52171b060a09d6e5c31c6c5ab8f10b9e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.6.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:571c44a8b7b9f5a387f98f65a8d7d74e0b6e5c80300da757c2713ff02a67889a +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..e34125b1fe53f87265628380dd8aeca3d48faa1e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0009672619635239244, "fvu": 0.005885841324925423} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..81c0db89672a8a67c1d20b030bef503fd2ee7280 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8927f1715b356bd1ee052f4af526cc2897df7cc026a6b1ebe713b505f6cc6f49 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..db46cd5019ad50c323890e92ec8b0d02455f04c1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.004151101689785719, "dead_pct": 7.440476474585012e-05} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..40d2e56d8ca4c253ec560a4c5e152bca0d408c73 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.7.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87053b43aa02d4cd9a9003b46a2087461245017313ba76ffa9b4634b92957a7e +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fe1809f673b38d54aba4e80e84366b39d3def9a2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0056547620333731174, "fvu": 0.005204628687351942} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d359bad108f6645bc5b494384defe1fdcdae5069 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0561f38f81373d0ed64d9ff8307ad0b24c309428905933b0ae199419bec0d06a +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cae09629173bb0f8d368b0a7ff9b6b83f00eb1e7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.004474824760109186, "dead_pct": 0.0004464286030270159} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2c20602c527fe3e135abdfde14621ada586d3466 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.8.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74c8e128822e233c2a8b0a7424f6dfa89ab4c6d8cddeabb52d719c8bcfcdff52 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..95ea72df7eab990b18d7b4b28e892a845acbb990 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.006090089213103056, "dead_pct": 0.0030505952890962362} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..33e9af81a92b951af760fe018f92a6a07563c0dc --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95d2b1f969f37b42f326dcd8990891b86d85e17b6f6867f51c3b462fb0a5a86b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..9a6f583aca1e21b0652587d5f8094c04fa7d7325 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..df07d96ff36cb73268d240681587e1dc527bca0d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.0016369047807529569, "fvu": 0.0050203013233840466} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9b239840014a6e2a31581a27dc049c8d50f1aa73 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model.layers.9.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c0a7d788885543938ac985057626ef96bcd7feefa46d3efd0449d091055b7e0e +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model_config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a96bcf4f590864a5b571e20ff6dbb25da603f0e4 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm2_cs1_experiment_4.3", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/optimizer.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..9f2b8678417d18251445d5df93934f6123c09dbe --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2f3ee98b9fc864e6274fd7d0733f5a1629183f42a028c6a7758ca39f73bbda77 +size 1468809018 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/state.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..3a962306fc6df2ec4b0fd0212513d4871a47eb44 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=256/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:228f70e2760f8926f98c2eefbfb520564a087b1357f7c52a7d9432710ba29550 +size 3233272 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b5719aa582b440208e1a6c1bbdc8f9f71c078a69 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_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": ["model.layers.5.mlp", "model.layers.5.post_attention_layernorm", "model.layers.6.mlp", "model.layers.6.post_attention_layernorm", "model.layers.7.mlp", "model.layers.7.post_attention_layernorm", "model.layers.8.mlp", "model.layers.8.post_attention_layernorm", "model.layers.9.mlp", "model.layers.9.post_attention_layernorm", "model.layers.10.mlp", "model.layers.10.post_attention_layernorm", "model.layers.11.mlp", "model.layers.11.post_attention_layernorm", "model.layers.12.mlp", "model.layers.12.post_attention_layernorm", "model.layers.13.mlp", "model.layers.13.post_attention_layernorm", "model.layers.14.mlp", "model.layers.14.post_attention_layernorm", "model.layers.15.mlp", "model.layers.15.post_attention_layernorm", "model.layers.16.mlp", "model.layers.16.post_attention_layernorm", "model.layers.17.mlp", "model.layers.17.post_attention_layernorm", "model.layers.18.mlp", "model.layers.18.post_attention_layernorm", "model.layers.19.mlp", "model.layers.19.post_attention_layernorm"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32", "wandb_log_frequency": 1} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/lr_scheduler.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/lr_scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1619aeb0b307c4dbe1e57034770cbd69eb75c48b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/lr_scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c1a4ccded57e94e581be291e5bc725b8c8e883138bcad563998f929ce54ece4 +size 1652 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..46583d60bba46c144d9247e5eb2b94dd5133bf14 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.023567013442516327, "dead_pct": 0.3656250238418579} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7d40106c03a9d41ba6cfbb34448f5b26667025dd --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b50784fbbd6fc430ce2ba5e4183c5173847e2319e8ccfdbaaae51890598c6a3c +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..477b9a15ae8f88ee3f3ea6b4e061839f70ddbb1e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.019318576902151108, "dead_pct": 0.4125000238418579} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2d9307ec606548e6caab1d7f00c90220863bc18b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.10.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4783638ce8244eff79064af2d76fdf8bb5814df227114cd9680aa28e087001d +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fb019a5fc5311cf3e7ac0cb9da517d6567b4cb8c --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.42752978205680847, "fvu": 0.03326193988323212} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6aa095a5b3ab5107c87e9fbbc8637dc5cb3f9131 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f05de1edb67d70f8089413ad29b811779b82ec7e3030a441e6c15337590c5e78 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b9822f712e6a1242c4abf7a67286bf1755e89708 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.0239181499928236, "dead_pct": 0.4348214566707611} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dd575fe78bd82671c9e193f4a6482e260b7528d6 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.11.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13214533e74a563e023197ad5bc67392f8d6d82f37c30e166562c3d80360162b +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..90df02c89d4c0c2bcdcb9ee14f21e7a8abe96d4b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.03228722885251045, "dead_pct": 0.3661458492279053} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5bd9b230d9d2f4d82c9a958206dc8215429987e1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19fa84120ec02f34f0f0c849ad82411d3ad61acfd8a01dfd2ebcbb8a4c931bc0 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..873be38c0204165678a82d884e11472bdcce5f5e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.0242347102612257, "dead_pct": 0.4541666805744171} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7f5c73a0fed3a8ea9b1442720426e3f51c3d06e5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.12.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9cc0623f5cd6befaa736651fec56973ee0ebfde1c95248cf55afd906d26da977 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9954a76bf3564e3d6704ce453941f4d56e362a80 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.3708333373069763, "fvu": 0.03183581680059433} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d03a4104784d47e4dce179a1dfbc9f7f52bf834a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:967451ece6e211966690eddfda771c4af36d1e08a4bd9df591ecbd06fe1fe218 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..84e3b4ee6b05379deadd0a990c91545a5c9d7e92 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.026903148740530014, "dead_pct": 0.4691964387893677} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4ca2e68224bd2b0babeee41b31ac6e066c441008 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.13.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d190207fe0e9f8f4bb079b9034f24d65ed4c3ef8fc7b4666bb4322848aaca30 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1ab4dbae3c788246686c550142f6b0319822b933 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.04365876317024231, "dead_pct": 0.2913690507411957} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..27dd5f829e7aa8a62f2782b619eb9c38181181c0 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16b78990f7a6dbd381ae2c0fbaac3b5cb62b3031675b93053bef6698684417db +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8f5f4be86a304fc43012ee2ed246009be55e9879 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.4519345462322235, "fvu": 0.03127102553844452} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a4d40fa39f9eaced474a638aa9023335428ccd3a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.14.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50b2bdbef42eebe53f1678268221123f44aefe05c847772a6305527bea80a8bd +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..ab3edf8cd0eb307da6ca0f2b70f1012701f392e3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.26659226417541504, "fvu": 0.04883865267038345} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..14874eda5039bc4343e8029771f25e8dbc9f3cf5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3764230e2dfc4d6125963d7ab29f200c3dbeefcfd26f1b42a5a0b040df693d18 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..b19598ddf70ba9648e12f83edd49a04089df248a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.415401816368103, "fvu": 0.02928028255701065} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7029be7a384cbc58d184720ddda8952aac9dc2d2 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.15.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6eee7d68de8797b594c46c636e310f96f1e3149446decc278d810469451cc433 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..0b93ac6a1911febaff4d966ceab3311700df973a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.03987765312194824, "dead_pct": 0.2511904835700989} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..606174044308f3d695f7fd5eb47cc976b527554d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38c6dcf12a26dc3482f6ad7aeaf7a446ce47fe8aded661e93a8b5351a151ab5c +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..8b2280970c3668b4b3bee1538559d5b1ad6bafb6 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.44813990592956543, "fvu": 0.02691182866692543} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d944bc14692dc5ec97633dd55563d1d4625dd38b --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.16.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:959c4d1e0e6aaf3664db7d7556025eb98db4643f0e1e060f11cb38ccf09e052d +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..5b2cbd55941e4f6d7c1b17f1f2b1821dd6abaae5 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.3292410969734192, "fvu": 0.025398598983883858} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b9b42fe886a47766ee30ba9788dc356978b17f0c --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1dc9a280cb9c6bf35162a50e0e940a9834c709538c0772fe008d1b999bc62cb +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..45ccf3f97da4e08e47e1044bfd1335f86afa94f7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.424107164144516, "fvu": 0.0257593784481287} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e03c1ba62c13d82d7f8e8083c7d38f162bb9fd9d --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.17.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e438d2e257978a893807268494dde155e4369339edb085e0171793a7fd320d99 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..91b2ec7c22c2baa14165dff6af93ead705648841 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.03436405956745148, "dead_pct": 0.191592276096344} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..373949ef4bc6676bbae789b02a358dae05c16868 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6afb5f474139f9e77189b7f4e53a03e58922e7b1a21867e9229b05560b0e465a +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..501eaa15c33494e174c8ec43821e0a6f40a3804f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.3756696581840515, "fvu": 0.02519919164478779} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..be3050a26d4edb6856863218a399cfe91efd3748 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.18.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aac4d70ffa3c10efb81b8f10e9947c9ff3118cb9f9a3d1365dbc0e41e1fcdeb8 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..72be5f807d6fadc665bea26486c1f6c7ce79fcf1 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.24263393878936768, "fvu": 0.026424391195178032} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..da5528f2b846c5ff06a35dbf2b1e238df0a76f69 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:40c8a4cf6ae950b5c4c93e8a991c5b8e987523040a82e19c4412ca2296e5435f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..f5f0bb50e76c4c6038f3c9a59614cb04bed7e7cc --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.023896778002381325, "dead_pct": 0.35625001788139343} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8c0e0c4ddf68ccd2038fcf9054635dab6d919cab --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.19.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f31e6f4f28be5161565ceca080c0c7f80efeec9461110b87285c2fbf1e064796 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..21a6f33857b1c68ef2829cf573b54cef3b2c80b7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.0100932652130723, "dead_pct": 0.3805059790611267} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4daa0cfc200352b1eb3f8a9ef00b6e70f4b2e483 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:449395d21f7479c1d6561a4789dde84a71476ad8a6e72fa6c641996873c55063 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fc01cb0394d576a5de026840749c4992cd3b9142 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.4125000238418579, "fvu": 0.010461684316396713} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6bc65658520b7bbf4c8191d2b6d1719215885010 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.5.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:438b77f9077d554b355101210bfcccbefc6c26948d2870f5542f4a10d523f029 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..932ad1b661dcedcdbb9f60d8537437dcb178d526 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.014470926485955715, "dead_pct": 0.3492559790611267} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..871fa8837d7d9729acd9c1983e24cbe1742a110a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f27aced9b8a6d5c47b58cfefa01ca3a49d6feb6acaf241876a7d26a3d566a8cf +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..22c747071968402d53dbe4593bfc8d57b1c78cc3 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.010844578966498375, "dead_pct": 0.41897323727607727} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e21ce4b6bbf09ae9287a6eee1399218eba20dcba --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.6.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dbb33d6bfaef919f0be42c9c14ac41e68e97d4b3332bf2b8bb0581319c78a36e +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..9479b53c54badd73137c3a030af77b07447dfc55 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.01629028469324112, "dead_pct": 0.30252978205680847} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..579f42f34198a0f649f95d00248d8af798358066 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65a7dba4b9e1a893acce85477f111f29fe577277a824c5feabe52a82f008439c +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..cdb738d738eda492d4c1f5d7c1c6a9152a962e09 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.010635700076818466, "dead_pct": 0.3704613149166107} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f76a267ef8d556cf9392f3cd9f6a0df2bad4d89e --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.7.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a86e5235c1ee5636b9f487c24f4c141ed84f38e63b8b5048765eaab527442a2 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..1be1c945b18b4b22517ff3994b6083bd92349367 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.018410416319966316, "dead_pct": 0.3533482253551483} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c437628480f50f032f78128d8eaeae09fee32903 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c08b71737dd447d4eb70253c9d07923e42de0861baba7b8ff40e0bd067c8127f +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..c7119af0ef390ead3e9ffff39f99fa1122d75da7 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.37968751788139343, "fvu": 0.012711661867797375} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e03daf3394d8629e6d051d0da1d1e5f55c827f1a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.8.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d2b0a9af86bfbda09445ddfed7e2381f852cb521ca0f71f342a4eb741d8c621 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.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": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..fd6c76c66f353cdc9d6a130cc0665ff46fd2dc78 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/metrics.json @@ -0,0 +1 @@ +{"fvu": 0.023684771731495857, "dead_pct": 0.3076637089252472} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8933b0b203894702cf740ba6491bcd9dcc78976f --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.mlp/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04028a0f3439217d46b13a47e889e7490eb8448584aa489fe12210602307a7f7 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/cfg.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/cfg.json new file mode 100644 index 0000000000000000000000000000000000000000..96efbbf9aad2dd40f6d65d2fcaf721b15e9a748a --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/cfg.json @@ -0,0 +1 @@ +{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 896} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/metrics.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/metrics.json new file mode 100644 index 0000000000000000000000000000000000000000..401f5d7e07a61feb0cfdc64eb8d9915ad3200bf8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/metrics.json @@ -0,0 +1 @@ +{"dead_pct": 0.38370537757873535, "fvu": 0.015597653575241566} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/sae.safetensors b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/sae.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..33443fee82babc78ae0cc5acf035c914fa642fd8 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model.layers.9.post_attention_layernorm/sae.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f5f0d0d3927cd1c7db87308facd6b16c909119f9af5187fe71d110f30899987 +size 96395592 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model_config.json b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a96bcf4f590864a5b571e20ff6dbb25da603f0e4 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/model_config.json @@ -0,0 +1 @@ +{"model_name": "withmartian/sql_interp_bm2_cs1_experiment_4.3", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm2_cs1_experiment_4.3_syn=True"} \ No newline at end of file diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/optimizer.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..e4757de72a271f99859a33286da1d9024d6c1ccd --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6c0fffc2a7b86c9b8971c6b586a6fa79a6aaef9aeb9d77fd46236d5d76827bb +size 1468809018 diff --git a/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/state.pt b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/state.pt new file mode 100644 index 0000000000000000000000000000000000000000..c0b80e8257e8f854564f8da8af57197681e2e059 --- /dev/null +++ b/saes_sql_interp_bm2_cs1_experiment_4.3_syn=True/k=32/state.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bc616c62157a0ce7e0969f9e863a56afebdfd35a03a1e2b75f73ae427c6d249b +size 3233272