Error when creating more than one split

#2
by mrdbourke - opened

Hello,

My loading script food_vision_199_classes.py works when loading images for the training dataset, however, when I try to use multiple splits, it fails.

I'm running the testing code:

datasets-cli test food_vision_199_classes.py --save_info --all_configs

  • Datasets version: 2.7.1
  • Python version: 3.8.13

I've narrowed the error down to how many splits I return in the _split_generators() method.

For example, only returning one datasets.SplitGenerator() works:

   def _split_generators(self, dl_manager):
        """
        This function returns the logic to split the dataset into different splits as well as labels.
        """
        annotations_csv = dl_manager.download("annotations_with_links.csv")
        print(annotations_csv)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "annotations": annotations_csv,
                    "split": "train"
                }
            ),
        ##### Commented out TEST split #####
        #     datasets.SplitGenerator(
        #         name=datasets.Split.TEST,
        #         gen_kwargs={
        #             "annotations": annotations_csv,
        #             "split": "test"
        #         }
        #     )
        ]

But when I try two splits, it doesn't work:

    def _split_generators(self, dl_manager):
        """
        This function returns the logic to split the dataset into different splits as well as labels.
        """
        annotations_csv = dl_manager.download("annotations_with_links.csv")
        print(annotations_csv)

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "annotations": annotations_csv,
                    "split": "train"
                }
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "annotations": annotations_csv,
                    "split": "test"
                }
            )
        ]

I get the following error (after running datasets-cli test food_vision_199_classes.py --save_info --all_configs):

Datasets version: 2.7.1
Verbosity level: 10
Testing builder 'default' (1/1)
Downloading and preparing dataset food_vision_199_classes/default to /home/daniel/.cache/huggingface/datasets/food_vision_199_classes/default/0.0.0/95cc7959e7881588f10aab90899c14edc40197b884aef48406145b08756412df...
annotations_with_links.csv
name: None                                                                                                         
Traceback (most recent call last):
  File "/home/daniel/code/pytorch/env/bin/datasets-cli", line 8, in <module>
    sys.exit(main())
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/commands/datasets_cli.py", line 39, in main
    service.run()
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/commands/test.py", line 141, in run
    builder.download_and_prepare(
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/builder.py", line 822, in download_and_prepare
    self._download_and_prepare(
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/builder.py", line 1555, in _download_and_prepare
    super()._download_and_prepare(
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/builder.py", line 913, in _download_and_prepare
    self._prepare_split(split_generator, **prepare_split_kwargs)
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/builder.py", line 1356, in _prepare_split
    split_info = self.info.splits[split_generator.name]
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/splits.py", line 525, in __getitem__
    instructions = make_file_instructions(
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/arrow_reader.py", line 111, in make_file_instructions
    name2filenames = {
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/arrow_reader.py", line 112, in <dictcomp>
    info.name: filenames_for_dataset_split(
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/naming.py", line 78, in filenames_for_dataset_split
    prefix = filename_prefix_for_split(dataset_name, split)
  File "/home/daniel/code/pytorch/env/lib/python3.8/site-packages/datasets/naming.py", line 57, in filename_prefix_for_split
    if os.path.basename(name) != name:
  File "/home/daniel/code/pytorch/env/lib/python3.8/posixpath.py", line 143, in basename
    p = os.fspath(p)
TypeError: expected str, bytes or os.PathLike object, not NoneType

See the full code for food_vision_199_classes.py here:
https://huggingface.co/datasets/mrdbourke/food_vision_199_classes/blob/main/food_vision_199_classes.py

Any help would be greatly appreciated!

cc @albertvillanova @lhoestq @severo .

Hi @mrdbourke ! Your dataset doesn't pass checks for the information about splits that is specified in README.md: it contains only train split but now you also have test. This seems to be a bug actually.
To avoid this, you can remove these lines from your README.md:

  splits:
  - name: train
    num_bytes: 2973286
    num_examples: 19747

Also, you can use ImageFolder the data structure you have: https://huggingface.co/docs/datasets/v2.7.1/en/image_dataset#imagefolder. That way you don't need to write any scripts. For your case, you don't even need .csv metadata files as you only have labels :)

Sign up or log in to comment