Data loader produces full test and validation split as well

opusparcus.py

@@ -236,7 +236,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
                         "quality": data["quality"],
                     }
         else:
-            keep_all = (split == "full-validation" || split == "full-test")
+            keep_all = (split == "full-validation" or split == "full-test")
             with open(filepath, encoding="utf-8") as f:
                 for id_, row in enumerate(f):
                     data = json.loads(row)