add loader

cs_fleurs.py

@@ -0,0 +1,74 @@
+import json
+import os
+from datasets import DatasetInfo, GeneratorBasedBuilder, BuilderConfig, SplitGenerator, Split, Features, Value
+
+class CSFLEURSConfig(BuilderConfig):
+    """BuilderConfig for a subset of CS-FLEURS."""
+    def __init__(self, subset_name=None, **kwargs):
+        super().__init__(**kwargs)
+        self.subset_name = subset_name  # e.g., "read", "xtts", "mms"
+
+class CSFLEURS(GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        CSFLEURSConfig(name="default", subset_name="all", description="All splits"),
+        CSFLEURSConfig(name="read", subset_name="read", description="Read subset"),
+        CSFLEURSConfig(name="xtts", subset_name="xtts", description="XTTS subset"),
+        CSFLEURSConfig(name="mms", subset_name="mms", description="MMS subset"),
+    ]
+
+    def _info(self):
+        return DatasetInfo(
+            description="CS-FLEURS: Code-switched speech dataset",
+            features=Features({
+                "id": Value("string"),
+                "file_name": Value("string"),
+                "text": Value("string"),
+                "duration": Value("float"),
+                "language": Value("string"),
+                "speaker": Value("string"),
+            }),
+            supervised_keys=None,
+        )
+
+    def _split_generators(self, dl_manager):
+        """Dynamically generate splits from subdirectories."""
+        base_path = os.path.abspath(".")
+        subset_dir = os.path.join(base_path, self.config.subset_name)
+
+        if not os.path.exists(subset_dir):
+            raise FileNotFoundError(f"Subset directory not found: {subset_dir}")
+
+        splits = []
+        for split_name in os.listdir(subset_dir):
+            split_path = os.path.join(subset_dir, split_name)
+            if os.path.isdir(split_path):
+                # Map folder names to Hugging Face Split enum if possible
+                hf_split_name = Split.TRAIN if split_name == "train" else Split.TEST
+                splits.append(
+                    SplitGenerator(
+                        name=hf_split_name if split_name in ["train", "test"] else split_name,
+                        gen_kwargs={"data_dir": split_path}
+                    )
+                )
+
+        return splits
+
+    def _generate_examples(self, data_dir):
+        """Yields examples from metadata.jsonl in the given directory."""
+        metadata_path = os.path.join(data_dir, "metadata.jsonl")
+        if not os.path.exists(metadata_path):
+            raise FileNotFoundError(f"Metadata file not found: {metadata_path}")
+
+        with open(metadata_path, "r", encoding="utf-8") as f:
+            for line in f:
+                entry = json.loads(line)
+                # Make file_name absolute
+                file_path = os.path.join(data_dir, entry["file_name"]) if not os.path.isabs(entry["file_name"]) else entry["file_name"]
+                yield entry["id"], {
+                    "id": entry["id"],
+                    "file_name": file_path,
+                    "text": entry["text"],
+                    "duration": float(entry["duration"]),
+                    "language": entry["language"],
+                    "speaker": entry["speaker"],
+                }