give a specific exception for common error

libs/libcommon/src/libcommon/exceptions.py

@@ -100,6 +100,7 @@ def as_response(self) -> ErrorResponse:
     "ExternalServerError",
     "FeaturesError",
     "FeaturesResponseEmptyError",
+    "FileFormatMismatchBetweenSplits",
     "FileSystemError",
     "HfHubError",
     "InfoError",
@@ -331,6 +332,13 @@ def __init__(self, message: str, cause: Optional[BaseException] = None):
         super().__init__(message, HTTPStatus.INTERNAL_SERVER_ERROR, "FeaturesResponseEmptyError", cause, True)
 
 
+class FileFormatMismatchBetweenSplits(CacheableError):
+    """Couldn't infer the same data file format for all splits."""
+
+    def __init__(self, message: str, cause: Optional[BaseException] = None):
+        super().__init__(message, HTTPStatus.INTERNAL_SERVER_ERROR, "FileFormatMismatchBetweenSplits", cause, False)
+
+
 class FileSystemError(CacheableError):
     """An error happen reading from File System."""
 

services/worker/src/worker/job_runners/dataset/config_names.py

@@ -21,6 +21,7 @@
     DatasetWithScriptNotSupportedError,
     DatasetWithTooManyConfigsError,
     EmptyDatasetError,
+    FileFormatMismatchBetweenSplits,
 )
 
 from worker.dtos import CompleteJobResult, ConfigNameItem, DatasetConfigNamesResponse
@@ -114,6 +115,8 @@ def compute_config_names_response(
     except ValueError as err:
         if "trust_remote_code" in str(err):
             raise DatasetWithScriptNotSupportedError from err
+        if "Couldn't infer the same data file format for all splits" in str(err):
+            raise FileFormatMismatchBetweenSplits() from err
         raise ConfigNamesError("Cannot get the config names for the dataset.", cause=err) from err
     except ImportError as err:
         # this should only happen if the dataset is in the allow list, which should soon disappear