updated loading script

xnli_cm_sample.py

@@ -49,6 +49,10 @@ _LICENSE = ""
 # The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
 # This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
 _URLS = {
+    "monolingual": 
+        {"train" : "data/monolingual/train.tsv",
+         "test" : "data/monolingual/test.tsv"
+         },
     "fr_ec": 
         {"train" : "data/fr_ec/train.tsv",
          "test" : "data/fr_ec/test.tsv"
@@ -88,13 +92,14 @@ class XnliCodeMixedSampled(datasets.GeneratorBasedBuilder):
     # data = datasets.load_dataset('my_dataset', 'second_domain')
 
     BUILDER_CONFIGS = [
+        datasets.BuilderConfig("monolingual", version=VERSION, description="Monolingual dataset"),
         datasets.BuilderConfig("fr_ec", version=VERSION, description="French-English code mixed dataset using Equivalence Constraint Theory"),
         datasets.BuilderConfig("fr_ml", version=VERSION, description="French-English code mixed dataset using Matrix Language Theory"),
         datasets.BuilderConfig("de_ec", version=VERSION, description="German-English code mixed dataset using Equivalence Constraint Theory"),
         datasets.BuilderConfig("de_ml", version=VERSION, description="German-English code mixed dataset using Matrix Language Theory"),
     ]
 
-    DEFAULT_CONFIG_NAME = "fr_ec"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+    DEFAULT_CONFIG_NAME = "monolingual"  # It's not mandatory to have a default configuration. Just use one if it make sense.
 
     def _info(self):
         # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset