Add data_loader

data_loader.py

@@ -0,0 +1,29 @@
+from datasets import DatasetInfo, GeneratorBasedBuilder, SplitGenerator, Split
+import datasets
+
+class SinhalaPerplexityDataset(datasets.GeneratorBasedBuilder):
+    def _info(self):
+        return DatasetInfo(
+            features=datasets.Features({
+                "sinhala_unicode": datasets.Value("string"),
+                "sinhala_romanized": datasets.Value("string"),
+            }),
+            description="Sinhala Perplexity Test Dataset with Unicode and Romanized Sinhala sentences.",
+            supervised_keys=None,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_path = dl_manager.download_and_extract("data/sinhala.csv")
+        return [
+            SplitGenerator(name="data", gen_kwargs={"filepath": data_path}),
+        ]
+
+    def _generate_examples(self, filepath):
+        import csv
+        with open(filepath, encoding="utf-8") as f:
+            reader = csv.DictReader(f)
+            for idx, row in enumerate(reader):
+                yield idx, {
+                    "sinhala_unicode": row["sinhala_unicode"],
+                    "sinhala_romanized": row["sinhala_romanized"],
+                }