Snap

clustering_files.py

@@ -48,7 +48,7 @@ class ClusteringFiles(datasets.GeneratorBasedBuilder):
         features = datasets.Features(
             {
                 "filename": datasets.Value("string"),
-                "segments": [
+                "segments": datasets.Sequence(
                     {
                         "segment_id": datasets.Value("int32"),
                         "speaker": datasets.Value("string"),
@@ -60,7 +60,7 @@ class ClusteringFiles(datasets.GeneratorBasedBuilder):
                         "readable_end": datasets.Value("string"),
                         "vp": datasets.Sequence(datasets.Value("float32")),
                     }
-                ]
+                )
             })
         return datasets.DatasetInfo(
             description=description,
@@ -100,8 +100,12 @@ class ClusteringFiles(datasets.GeneratorBasedBuilder):
                 "end": row['end'],
                 "readable_start": row['readable_start'],
                 "readable_end": row['readable_end'],
-                "vp": np.asarray(row["vp"], dtype=np.float32)
+                "vp": row["vp"]  # Keep as numpy array
             })
 
         for idx, file_data in enumerate(files.values()):
             yield idx, file_data
+
+# Ensure numpy arrays are properly handled by datasets library
+datasets.config.NUMPY_ARRAYS_AVAILABLE = True
+datasets.config.NUMPY_DTYPE_TO_PYARROW_DTYPE[np.dtype("float32")] = datasets.Value("float32")