Upload trainer_0/config.json with huggingface_hub

trainer_0/config.json

@@ -2,8 +2,8 @@
     "trainer": {
         "trainer_class": "TopKTrainer",
         "dict_class": "AutoEncoderTopK",
-        "lr": 0.0002,
-        "steps": 12,
+        "lr": 0.000282842712474619,
+        "steps": 61035,
         "auxk_alpha": 0.03125,
         "warmup_steps": 0,
         "decay_start": null,
@@ -11,7 +11,7 @@
         "threshold_start_step": 1000,
         "seed": null,
         "activation_dim": 6144,
-        "dict_size": 16384,
+        "dict_size": 8192,
         "k": 128,
         "device": "cuda:0",
         "layer": 4,
@@ -23,9 +23,9 @@
         "n_models": 12,
         "d_submodule": 512,
         "io": "out",
-        "n_ctxs": 512,
+        "n_ctxs": 1024,
         "ctx_len": 128,
-        "refresh_batch_size": 256,
+        "refresh_batch_size": 512,
         "out_batch_size": 8192,
         "device": "cuda:0",
         "rescale_acts": true