Datasets:
Dataset Viewer issue: TooBigContentError
The dataset viewer is not working.
Error details:
Error code: TooBigContentError
I realize this may be a bigger issue that can't be fixed as easily (based on this issue).
However, it also won't read the CSV paired with the smaller images for the resized subset and I'm not sure why it's not combining them. TIA for your help!
cc @albertvillanova @lhoestq @severo .)
I wonder if I could maybe implement a workaround for the larger images to at least get thumbnails aligned with the metadata by using the resolve URL for the full size images in a file_url column.
"However, it also won't read the CSV paired with the smaller images for the resized subset and I'm not sure why it's not combining them. TIA for your help!":
I'm not 100% sure, but maybe replacing:
- config_name: resized_images
data_files:
- BeetleMeasurements_resized.csv
- beetle_images_resized/*.jpg
with
- config_name: resized_images
data_files:
- beetle_images_resized/*
and moving BeetleMeasurements_resized.csv to beetle_images_resized/metadata.csv.
cc @albertvillanova and @polinaeterna for confirmation.
I wonder if I could maybe implement a workaround for the larger images to at least get thumbnails aligned with the metadata by using the resolve URL for the full size images in a file_url column.
yes, indeed, it could work (note that users that access the page might have to download 100's of MB... )
@severo
, I tried moving the metadata file (renaming it to metadata.csv and adjusting the filepaths in file_name to reflect the new placement), but I still can't get the metadata to align with the images in the resized viewer.
weird that the metadata.csv file is ignored. Maaaybe put everything in a train/ subdirectory? See https://huggingface.co/collections/datasets-examples/image-dataset-6568e7cf28639db76eb92d65 + https://huggingface.co/docs/hub/datasets-image. (sorry I don't know in-depth how image folders work). cc @lhoestq in case it's something obvious I missed
weird that the metadata.csv file is ignored. Maaaybe put everything in a train/ subdirectory? See https://huggingface.co/collections/datasets-examples/image-dataset-6568e7cf28639db76eb92d65 + https://huggingface.co/docs/hub/datasets-image. (sorry I don't know in-depth how image folders work). cc @lhoestq in case it's something obvious I missed
That shouldn't matter. My initial setup mirrored our Rare Species Dataset with the configs pointing to the appropriate files (as I also did with our butterfly dataset that has custom named CSVs, but the images are displayed through URLs instead of uploaded).