The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 81, in _split_generators
first_examples = list(islice(pipeline, self.NUM_EXAMPLES_FOR_FEATURES_INFERENCE))
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 32, in _get_pipeline_from_tar
fs: fsspec.AbstractFileSystem = fsspec.filesystem("memory")
~~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 302, in filesystem
cls = get_filesystem_class(protocol)
File "/usr/local/lib/python3.14/site-packages/fsspec/registry.py", line 239, in get_filesystem_class
raise ValueError(f"Protocol not known: {protocol}")
ValueError: Protocol not known: memory
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 71, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Typst Image Dataset
This dataset was generated with a fork of tex2typ and the hoang-quoc-trung/fusion-image-to-latex-datasets dataset, which itself is a compilation of LaTeX labels and images of equations.
The hoang-quoc-trung dataset is difficult to work with in that it has the image data stored in a large compressed RAR archive, which does not permit efficient random read access. Additionally, it appears to have a larger number of corrupted filenames inside the archive, which has been mended in this dataset.
This dataset instead opts to use a WebDataset for convenient and efficient storage of the image files and associated metadata.
The code used to generate this dataset can be found at here: https://github.com/JeppeKlitgaard/DTU-02456-Deep-Learning-Project (this is currently private but should be released after examination. If this is not the case prod me at huggingface@jeppe.science).
Note that the data follows the WebDataset convention, but notably does not follow it's suggestions when it comes to image file extensions (or file extensions in general) inside the archive. This is due to the way in which HuggingFace's datasets library implements
WebDatasets, which makes it difficult to rejoin the image columns without going through a needlessly expensive map call.
As a result, however, datasets is not able to automatically detect the feature types, instead we may define them explicitly:
features = Features(
{
"input_image": Image(),
"latex": Value("string"),
"typst": Value("string"),
"typst_image": Image(),
"metadata.json": {
"image_type": ClassLabel(names=["handwritten", "printed"]),
"image_extension": ClassLabel(names=["png", "jpg", "bmp", "dvi"]),
}
}
)
Note that typst_image is an image of the typst string that has been rendered using Typst 0.14 in the PNG format with 120.0 ppi using template and default fonts:
#set page(width: auto, height: auto, margin: 0pt)
$
{typst_code}
$
These are included such that a cost function (such as intersection-over-union) can be used in training by rendering the predicted Typst source and comparing it to the known good picture. If the Typst version and ppi setting is kept the same, any equivalent source code should produce exactly the same pixels.
- Downloads last month
- 266