Abdullah
commited on
Uploading saes for ['transformer.h.0.attn', 'transformer.h.0.mlp', 'transformer.h.1.attn', 'transformer.h.1.mlp'] and sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32
Browse files- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/config.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/lr_scheduler.pt +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/model_config.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/optimizer.pt +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/state.pt +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/cfg.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/metrics.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/sae.safetensors +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/cfg.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/metrics.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/sae.safetensors +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/cfg.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/metrics.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/sae.safetensors +3 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/cfg.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/metrics.json +1 -0
- saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/sae.safetensors +3 -0
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"sae": {"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false}, "batch_size": 16, "grad_acc_steps": 1, "micro_acc_steps": 1, "lr": null, "lr_warmup_steps": 1000, "auxk_alpha": 0.0, "dead_feature_threshold": 10000000, "hookpoints": ["transformer.h.0.attn", "transformer.h.0.mlp", "transformer.h.1.attn", "transformer.h.1.mlp"], "init_seeds": [0], "layers": [], "layer_stride": 1, "transcode": false, "distribute_modules": false, "save_every": 100000, "log_to_wandb": true, "run_name": "saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32", "wandb_log_frequency": 1}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/lr_scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:81ec4a68d15406a4705b860cc8ccb0cda3b0bf3d809162715b62225325b749eb
|
| 3 |
+
size 1483
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/model_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"model_name": "withmartian/sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1", "dataset_name": "withmartian/cs4_dataset_synonyms", "model_abbrev": "saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True"}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2cc8c93d4c497d1fabbe67f05b70d065e30fe6426b551e70747ae2b0016ec8e4
|
| 3 |
+
size 1007171903
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/state.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29aad59d85f341825e4d10a1f104113baadbfb142c9b3e02f1d97d7f026e2e9a
|
| 3 |
+
size 493769
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/cfg.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/metrics.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"dead_pct": 0.44225263595581055, "fvu": 0.03268370404839516}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.attn/sae.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0c32237943a713b7a79565106b8e3a372eff402f087e973fbaafd622e276beb3
|
| 3 |
+
size 125894984
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/cfg.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/metrics.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"dead_pct": 0.0758463591337204, "fvu": 0.04444556310772896}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.0.mlp/sae.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6052653c7f8dc5db4963f9f8bb41561a2448fe8c7b661b57673f4e7e5c7cc26c
|
| 3 |
+
size 125894984
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/cfg.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/metrics.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"dead_pct": 0.5193359851837158, "fvu": 0.04987297207117081}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.attn/sae.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fb04903cff6c734d855ab50723280e117793f6407d7a8d9f71b8f6c9d927e0a4
|
| 3 |
+
size 125894984
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/cfg.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"expansion_factor": 15, "normalize_decoder": true, "num_latents": 0, "k": 32, "multi_topk": false, "skip_connection": false, "d_in": 1024}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/metrics.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"dead_pct": 0.14414063096046448, "fvu": 0.06485763192176819}
|
saes_sft_sql_interp_TinyStories-2Layers-33M_cs4_dataset_synonyms_experiment_1.1_syn=True/k=32/transformer.h.1.mlp/sae.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:04a95612702cdde9dcad86a2c916dbf0deddea849da35b842d1071878f062271
|
| 3 |
+
size 125894984
|