Upload sparsity_coefficient_0.000131_20240314001404151327_config.json with huggingface_hub

sparsity_coefficient_0.000131_20240314001404151327_config.json

@@ -0,0 +1,36 @@
+{
+    "seed": 42,
+    "model_name_or_path": "EleutherAI/pythia-70m-deduped",
+    "revision": "",
+    "hook_point": "gpt_neox.layers.3",
+    "dataset_name_or_path": "jbrinkma/pile-300k",
+    "activation_size": -1,
+    "add_bos_token": false,
+    "evaluation_batches": 10,
+    "expansion_factor": 4,
+    "b_dec_init_method": "",
+    "n_steps": -1,
+    "device": "cuda",
+    "batch_size": 32,
+    "ctx_length": 256,
+    "lr": 0.001,
+    "min_lr": 0.0,
+    "lr_warmup_steps": 5000,
+    "sparsity_coefficient": 0.00013142857142857143,
+    "evaluation_interval": 200,
+    "beta1": 0.9,
+    "beta2": 0.999,
+    "l1_sqrt": true,
+    "cos_sim_reg": false,
+    "cos_sim_alpha": 0.0,
+    "n_tokens_in_feature_cache": 500000.0,
+    "use_ghost_grads": false,
+    "output_dir": "outputs",
+    "cache_dir": "cache",
+    "checkpoint_interval": 200,
+    "use_wandb": true,
+    "wandb_entity": "jannikbrinkmann",
+    "wandb_project": "best-sae",
+    "wandb_name": "sparsity_coefficient_0.000131_20240314001404151327",
+    "wandb_group": "L1(Sqrt(Sqrt))"
+}