Offline dataset viewer

[yuvalkirstain]

Feature request

The dataset viewer feature is very nice. It enables to the user to easily view the dataset. However, when working for private companies we cannot always upload the dataset to the hub. Is there a way to create dataset viewer offline? I.e. to run a code that will open some kind of html or something that makes it easy to view the dataset.

Motivation

I want to easily view my dataset even when it is hosted locally.

Your contribution

N.A.

[severo]

Hi, thanks for the suggestion. It's not possible at the moment. The viewer is part of the Hub codebase and only works on public datasets. Also, it relies on Datasets Server, which prepares the data and provides an API to access the rows, size, etc.

If you're interested in hosting your data as a private dataset on the Hub, you might want to look at huggingface/dataset-viewer#39.

[ssuwelack]

Hi, we are building an offline dataset viewer: https://github.com/Renumics/spotlight
It supports many HF datasets, but currently you have to use it via Pandas:
df=ds.to_pandas()
spotlight.show(df)

Would love to hear from you if that works for your use case. If not, feel free to open an issue on the repo: https://github.com/Renumics/spotlight/issues

[yuvalkirstain]

@ssuwelack thank you! I will definitely try it out.

[severo]

Related issues:

• Standalone viewer dataset-viewer#213
• Opensource the Hub dataset viewer dataset-viewer#441
• Request to Share/Update Dataset Viewer Code #6014

[severo]

Closing for now, as developing and maintaining an offline viewer is not planned.

[severo]

@yuvalkirstain the dataset viewer is now available on private datasets for PRO users and Enterprise Hub orgs. Would it fit your needs?

[laubonghaudoi]

Hi @ssuwelack I tried loading a HF dataset with your viewer but got this error Renumics/spotlight#461 hope the team can help me on this. Thanks!