Upload SAEs from experiment 4.3
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/lr_scheduler.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/model_config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/optimizer.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/state.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/lr_scheduler.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/model_config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/optimizer.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/state.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/lr_scheduler.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/model_config.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/optimizer.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/state.pt +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/metrics.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors +3 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.mlp/cfg.json +1 -0
- saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.mlp/metrics.json +1 -0
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128", "wandb_log_frequency": 1}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/lr_scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:afafb111606b64199b463308d054f28acf71d5cca5174e9fd9fce0b0ab0a717c
|
3 |
+
size 1076
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/model_config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"model_name": "withmartian/sql_interp_bm1_cs1_experiment_1.10", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True"}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:84ac1e2a4b8279d0b3058c6ca4e4196dba16fd257331c187b67ede2fec3fff5b
|
3 |
+
size 255767204
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/state.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a612e8ae5676765040241e0d09a5e743af130e8d444b32987c09e9bf68d244b6
|
3 |
+
size 493312
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.18085938692092896, "fvu": 0.002518476452678442}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:aa3539bc8788ca8235e06606a9bfecfc247fc329863b899cdfa1130e50ce026f
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"fvu": 0.019294889643788338, "dead_pct": 0.0}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.0.mlp/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d91b236093cd5c4a8247ed11cede8fb71fe2414783345045688935a710f53e3e
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.22486980259418488, "fvu": 0.020067255944013596}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:266bead4788280baf24dbb0c88d6df9d77da99a2ac1831b010bdf6793c55de5d
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 128, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"fvu": 0.0174215380102396, "dead_pct": 0.01484375074505806}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=128/transformer.h.1.mlp/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cf7f5911846fdff382b39127c0b731012966fb6261af6d8e128c9cf6799b2455
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"sae": {"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16", "wandb_log_frequency": 1}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/lr_scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a19ec7a7d99a9517726d577d84860d99635fa1571577af8b6d7ba1c019ae8379
|
3 |
+
size 1076
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/model_config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"model_name": "withmartian/sql_interp_bm1_cs1_experiment_1.10", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True"}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:de72be0333f6c32c60298b6dfe17542989da4009b74f4cac9ce4aa480d1f281b
|
3 |
+
size 68243108
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/state.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b936b2797e1a1d9d822067e4f761d6771c4063ecfc5d241b7a4f387f66d43472
|
3 |
+
size 132864
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.6455078125, "fvu": 0.019297020509839058}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:363144ece15fafd7b196c17966e2e82861105406f3693a8ce89029f39869c2ce
|
3 |
+
size 33575240
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.245361328125, "fvu": 0.03865286707878113}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.0.mlp/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b2442484ede1e6d6e4eda07418e02f832ecd8c04e228a666f4bfaa814bc4d513
|
3 |
+
size 33575240
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.745849609375, "fvu": 0.05162069573998451}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:eb4f90210b67c975e41c6db0b94b2e9e4e4e3deea5f702c601352bbed36f36db
|
3 |
+
size 33575240
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 4, "normalize_decoder": true, "num_latents": 0, "k": 16, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.476806640625, "fvu": 0.040764667093753815}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=16/transformer.h.1.mlp/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:53ab0927cdca8b60d9d2761e93b7fb7b32588ce9b0de577d612117b42c165949
|
3 |
+
size 33575240
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256", "wandb_log_frequency": 1}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/lr_scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bd74b3fabf7af5e0d248711a1fa05fe261368df7b69f594226dfd4b050dd08b9
|
3 |
+
size 1076
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/model_config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"model_name": "withmartian/sql_interp_bm1_cs1_experiment_1.10", "dataset_name": "withmartian/cs1_dataset_synonyms", "model_abbrev": "saes_sql_interp_bm1_cs1_experiment_1.10_syn=True"}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:76be65f441881d9f9282674951baf9935e9dd6f4890b2c6c30c496d327570800
|
3 |
+
size 255767204
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/state.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:49a00b004d5cec4cd9f4163f4ba7d5a2db8308446a1d8131959f7589dd485fec
|
3 |
+
size 493312
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.21692709624767303, "fvu": 0.0020793573930859566}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:50c727d8005ec42bcedd7b4dabfff632e4b66048573c4523d75a91c7519ba66c
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"fvu": 0.013328234665095806, "dead_pct": 6.510417006211355e-05}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.0.mlp/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:87727975eeae167af6673f4b690af7383d895949b430b7ab0513881fb7b8a52f
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"dead_pct": 0.46438804268836975, "fvu": 0.008955885656177998}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.attn/sae.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:46d4315cefb51df71f2e6d7f1d1f2181f7dfec65a025982f4d4ba82bce5f7b4c
|
3 |
+
size 125894984
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.mlp/cfg.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 256, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sql_interp_bm1_cs1_experiment_1.10_syn=True/k=256/transformer.h.1.mlp/metrics.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"fvu": 0.011177952401340008, "dead_pct": 0.005924479570239782}
|