fix dataset loading by using the proper file extension `parquet`

Fixing loading the dataset as it is not working right now because the wrong file extension in the dataset config

README.md

@@ -21,7 +21,7 @@ configs:
 - config_name: default
   data_files:
   - split: train
-    path: "*.jsonl"
+    path: "data/*.parquet"
 ---
 
 # ArabicText-Large: High-Quality Arabic Corpus for LLM Training