Data loader produces full test and validation split as well

opusparcus.py

@@ -47,8 +47,6 @@ _LICENSE = ""
 _URLs = {
     "validation": "validation.jsonl",
     "test": "test.jsonl",
-    "full-validation": "validation.jsonl",
-    "full-test": "test.jsonl"
     # NB: the "train" split file is defined dynamically inside the `_split_generators` method
 }
 
@@ -172,7 +170,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
                 },
             ),
             datasets.SplitGenerator(
-                name="full-test",
+                name="test.full",
                 # These kwargs will be passed to _generate_examples
                 gen_kwargs={
                     "lang": self.config.lang,
@@ -182,7 +180,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
                 },
             ),
             datasets.SplitGenerator(
-                name="full-validation",
+                name="validation.full",
                 # These kwargs will be passed to _generate_examples
                 gen_kwargs={
                     "lang": self.config.lang,
@@ -236,7 +234,7 @@ class Opusparcus(datasets.GeneratorBasedBuilder):
                         "quality": data["quality"],
                     }
         else:
-            keep_all = (split == "full-validation" or split == "full-test")
+            keep_all = (split == "validation.full" or split == "test.full")
             with open(filepath, encoding="utf-8") as f:
                 for id_, row in enumerate(f):
                     data = json.loads(row)