The Dataset Preview has been disabled on this dataset


what are the prerequisites to have dataset preview visible on the dataset page?

I generated .parquet files using pyarrow and uploaded it using git push from CLI and I don’t see a preview.

Is the preview only supported when using push_to_hub method, or I am missing script file in my repo and that’s why the preview is not generated? That’s generally a more broad question about how the preview generation works under the hood. There is nothing in the docs about how it works and what are the requirements.


Hi! Is the dataset in question private? If I’m not mistaken, the preview only works for public datasets, but we plan to extend support to private datasets soon (cc @severo).

1 Like

Hi Piotr,

uploading a .parquet file to a repo should be enough to have the dataset viewer work. See julien-c/impressionists · Datasets at Hugging Face for example.
As mentioned by @mariosasko, the dataset viewer is not available for the private datasets.

In any case, we are working on making both the docs and the error messages clearer to make it easier to solve any potential problem with the dataset server.

Do you have a particular repo we could review to understand what is occurring?

1 Like

Thank you @mariosasko @severo ,

that’s exactly the reason why the preview was not available. The repository was private.

I also had to add dataset_infos.json file because one of the column contain bytes and path to Image object:

  "podbilabs--XYZ": {
    "description": "",
    "citation": "",
    "homepage": "",
    "license": "",
    "features": {
      "image": {
        "decode": true,
        "id": null,
        "_type": "Image"
      "ground_truth": {
        "dtype": "string",
        "id": null,
        "_type": "Value"
    "post_processed": null,
    "supervised_keys": null,
    "task_templates": null,
    "version": {
      "version_str": "0.0.0",
      "description": null,
      "major": 0,
      "minor": 0,
      "patch": 0
1 Like