Update dataset.py

dataset.py

@@ -4,7 +4,6 @@ import datasets
 import csv
 import pyarrow as pa
 
-# Default data files for the Hub viewer (when no data_files passed)
 DEFAULT_DATA_FILES = {
     "Thresholding": {
         "train": "train/train.csv",
@@ -37,32 +36,18 @@ class MyDataset(GeneratorBasedBuilder):
     ]
 
     def _info(self):
-        # Let Hugging Face infer features from data (no hard-coded schema)
-        return DatasetInfo()
+        return DatasetInfo()  # no schema assumptions
 
     def _split_generators(self, dl_manager):
-        # Try to get data_files passed in the config
-        data_files = getattr(self.config, "data_files", None)
-
-        # Fallback to default if not passed
-        if not data_files:
-            if self.config.name not in DEFAULT_DATA_FILES:
-                raise ValueError(f"No default files for config {self.config.name}")
-            data_files = [
-                {"split": k, "path": v}
-                for k, v in DEFAULT_DATA_FILES[self.config.name].items()
-            ]
-
-        def get_path(split_name, fallback=None):
-            for entry in data_files:
-                if entry["split"] == split_name:
-                    return entry["path"]
-            return fallback
+        # Try default files if config has no data_files
+        file_dict = DEFAULT_DATA_FILES.get(self.config.name)
+        if not file_dict:
+            raise ValueError(f"No default data files defined for config: {self.config.name}")
 
         return [
-            SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": get_path("train")}),
-            SplitGenerator(name=Split.VALIDATION, gen_kwargs={"filepath": get_path("dev") or get_path("val")}),
-            SplitGenerator(name=Split.TEST, gen_kwargs={"filepath": get_path("test")}),
+            SplitGenerator(name=Split.TRAIN, gen_kwargs={"filepath": file_dict["train"]}),
+            SplitGenerator(name=Split.VALIDATION, gen_kwargs={"filepath": file_dict.get("val") or file_dict.get("dev")}),
+            SplitGenerator(name=Split.TEST, gen_kwargs={"filepath": file_dict["test"]}),
         ]
 
     def _generate_examples(self, filepath):
@@ -70,8 +55,8 @@ class MyDataset(GeneratorBasedBuilder):
             try:
                 with open(filepath, "rb") as f:
                     table = pa.ipc.RecordBatchFileReader(f).read_all()
-            except pa.lib.ArrowInvalid:
-                raise ValueError(f"File {filepath} is not a valid Arrow file.")
+            except pa.lib.ArrowInvalid as e:
+                raise ValueError(f"Invalid Arrow file at {filepath}: {e}")
             data = table.to_pydict()
             for i in range(len(next(iter(data.values())))):
                 yield i, {k: data[k][i] for k in data}