Upload trainer_0/config.json with huggingface_hub
Browse files- trainer_0/config.json +7 -7
trainer_0/config.json
CHANGED
|
@@ -4,10 +4,10 @@
|
|
| 4 |
"trainer_class": "StandardTrainerAprilUpdate",
|
| 5 |
"activation_dim": 1792,
|
| 6 |
"dict_size": 57344,
|
| 7 |
-
"lr": 1e-
|
| 8 |
-
"l1_penalty": 0.
|
| 9 |
-
"warmup_steps":
|
| 10 |
-
"sparsity_warmup_steps":
|
| 11 |
"steps": 30517,
|
| 12 |
"decay_start": null,
|
| 13 |
"seed": null,
|
|
@@ -16,9 +16,9 @@
|
|
| 16 |
"lm_name": "blah",
|
| 17 |
"wandb_name": "StandardTrainerAprilUpdate",
|
| 18 |
"submodule_name": null,
|
| 19 |
-
"frac_features_shared": 0.
|
| 20 |
-
"shared_l1_penalty": 0.
|
| 21 |
-
"num_shared_features":
|
| 22 |
},
|
| 23 |
"buffer": {
|
| 24 |
"n_models": 2,
|
|
|
|
| 4 |
"trainer_class": "StandardTrainerAprilUpdate",
|
| 5 |
"activation_dim": 1792,
|
| 6 |
"dict_size": 57344,
|
| 7 |
+
"lr": 1e-05,
|
| 8 |
+
"l1_penalty": 0.1,
|
| 9 |
+
"warmup_steps": 20,
|
| 10 |
+
"sparsity_warmup_steps": 20,
|
| 11 |
"steps": 30517,
|
| 12 |
"decay_start": null,
|
| 13 |
"seed": null,
|
|
|
|
| 16 |
"lm_name": "blah",
|
| 17 |
"wandb_name": "StandardTrainerAprilUpdate",
|
| 18 |
"submodule_name": null,
|
| 19 |
+
"frac_features_shared": 0.04,
|
| 20 |
+
"shared_l1_penalty": 0.02,
|
| 21 |
+
"num_shared_features": 2293
|
| 22 |
},
|
| 23 |
"buffer": {
|
| 24 |
"n_models": 2,
|