Upload src/kfac_collector.py with huggingface_hub

src/kfac_collector.py

@@ -480,8 +480,8 @@ class KFACCollector:
 
 
 def create_dataloader(
-    dataset_name: str = "allenai/olmo-mix-1124",
-    dataset_config: str = None,  # Use default config which includes all subsets
+    dataset_name: str = "allenai/c4",
+    dataset_config: str = "en",  # English subset
     tokenizer = None,
     batch_size: int = 4,
     seq_length: int = 512,