make custom exceptions for FileNotFound in dataloader and `convert_…

…text_to_mds`

llmfoundry/data/finetuning/dataloader.py

@@ -25,6 +25,7 @@
 from llmfoundry.data.text_data import build_streams
 from llmfoundry.utils.config_utils import to_dict_container
 from llmfoundry.utils.exceptions import (
+    DatasetMissingFileError,
     MissingHuggingFaceURLSplitError,
     NotEnoughDatasetSamplesError,
 )
@@ -541,14 +542,7 @@ def _download_remote_hf_dataset(remote_path: str, split: str) -> str:
                 get_file(path=name, destination=destination, overwrite=True)
             except FileNotFoundError as e:
                 if extension == SUPPORTED_EXTENSIONS[-1]:
-                    files_searched = [
-                        f'{name}/{split}{ext}' for ext in SUPPORTED_EXTENSIONS
-                    ]
-                    raise FileNotFoundError(
-                        f'Could not find a file with any of ' + \
-                        f'the supported extensions: {SUPPORTED_EXTENSIONS}\n' + \
-                        f'at {files_searched}',
-                    ) from e
+                    raise DatasetMissingFileError(file_name=f"name/{split}") from e
                 else:
                     log.debug(
                         f'Could not find {name}, looking for another extension',

llmfoundry/utils/exceptions.py

@@ -4,6 +4,8 @@
 """Custom exceptions for the LLMFoundry."""
 from typing import Any, Dict, List, Literal, Optional, Union
 
+from llmfoundry.data.finetuning.tasks import SUPPORTED_EXTENSIONS
+
 __all__ = [
     'ALLOWED_RESPONSE_KEYS',
     'ALLOWED_PROMPT_KEYS',
@@ -354,3 +356,20 @@ class RunTimeoutError(InternalError):
     def __init__(self, timeout: int) -> None:
         message = f'Run timed out after {timeout} seconds.'
         super().__init__(message, timeout=timeout)
+
+
+class DatasetMissingFileError(UserError):
+    """Error thrown when a dataset cannot find a file."""
+    def __init__(self, file_name: List[str]) -> None:
+        message = "Could not find the file '{file_name}' with any of the supported extensions: "
+        message += ", ".join(SUPPORTED_EXTENSIONS) + '.'
+        message += " Please check your train / eval data and try again."
+        super().__init__(message, file_name=file_name)
+
+
+class DatasetInvalidFolderError(UserError):
+    """Error thrown when a dataset folder is invalid."""
+    def __init__(self, folder_path: str) -> None:
+        message = f"Could not find objects at the path '{folder_path}'. "
+        message += "Please check your `input_folder` and try again."
+        super().__init__(message, folder_path=folder_path)

scripts/data_prep/convert_text_to_mds.py

@@ -30,6 +30,7 @@
     merge_shard_groups,
 )
 from llmfoundry.utils.exceptions import (
+    DatasetInvalidFolderError,
     InputFolderMissingDataError,
     OutputFolderNotEmptyError,
 )
@@ -232,10 +233,13 @@ def get_object_names(input_folder: str) -> List[str]:
     object_store = maybe_create_object_store_from_uri(input_folder)
     if object_store is not None:
         _, _, folder_prefix = parse_uri(input_folder)
-        names = [
-            name for name in object_store.list_objects(folder_prefix)
-            if name.endswith('.txt')
-        ]
+        try:
+            names = [
+                name for name in object_store.list_objects(folder_prefix)
+                if name.endswith('.txt')
+            ]
+        except FileNotFoundError as e:
+            raise DatasetInvalidFolderError(input_folder) from e
     else:
         # input_folder is a local folder
         names = [