Dataset Viewer issue

#2
by cschell - opened
Owner

Hi,

The dataset viewer is not working.

Error details:

Error code:   DatasetWithScriptNotSupportedError

You seem have to disabled python scripts for this purpose – what would be the right way now to configure how subsets and samples should be loaded?

cc @albertvillanova @lhoestq @severo .

Hi. You can use the space: https://huggingface.co/spaces/albertvillanova/convert-dataset-to-parquet to convert your dataset to data-only. Or do it manually following the docs: https://huggingface.co/docs/hub/datasets-data-files-configuration
Please tell us how it worked for you!

Owner
β€’
edited Jan 15

Thanks for the quick reply!

Bildschirmfoto 2024-01-15 um 11.42.53.png
Unfortunately, it doesn't seem to have worked :-/

Are there any things I have to be aware of if my data consists of time series? Currently, each .parquet file in the repository (like here) contains one continuous sequence of frames (each row == 1 frame).

Edit: I am asking this, because when I initially tried getting this to work without a script, just the yaml config, load_dataset was very slow. Only when I implemented the loading script with ArrowBasedBuilder loading time became acceptable.

-chris

I think @albertvillanova will be able to give more details πŸ™

Owner

@albertvillanova , @severo hey, are there any updates? I have created a test-repo with just one subset of the data and just a datacard. Loading this repo with load_dataset("cschell/test-repo") works, but is very slow. Can I somehow modify the data card to improve loading time?

-chris

Sign up or log in to comment