Dataset preview doesn't working: "The split does not contain any rows."

[alkzar90]

Hi.

I created a dataset repository for one version of the CUB multimodal dataset, but the data preview doesn’t work. The dataset is downloaded and loaded correctly.

Any help would be great!

[severo]

It seems like streaming mode is not working on your dataset, ie: ds = load_dataset(dataset, split='train') works, but not ds = load_dataset(dataset, split='train', streaming=True).

And the dataset viewer is using the streaming mode to get the first 100 rows (more on this in the docs: Preview a dataset)

Maybe @albertvillanova or @lhoestq could give more hints on how to make the dataset streamable?

[albertvillanova]

As @severo points out, the current implementation does not support streaming mode.

We have opened an issue in the Community tab of the corresponding dataset repository. Please follow the thread there: alkzar90/CC6204-Hackaton-Cub-Dataset · Issue with dataset viewer

[alkzar90]

After the PR, now the streaming mode is working.

ds = load_dataset("alkzar90/CC6204-Hackaton-Cub-Dataset",
                  streaming=True)

Thanks for the support, every time I learn a bit more about using the dataset library with custom script!