Bad error message when loading private dataset

[patrickvonplaten]

Describe the bug

A pretty common behavior of an interaction between the Hub and datasets is the following.
An organization adds a dataset in private mode and wants to load it afterward.

from transformers import load_dataset

ds = load_dataset("NewT5/dummy_data", "dummy")

This command then fails with:

FileNotFoundError: Couldn't find a dataset script at /home/patrick/NewT5/dummy_data/dummy_data.py or any data file in the same directory. Couldn't find 'NewT5/dummy_data' on the Hugging Face Hub either: FileNotFoundError: Dataset 'NewT5/dummy_data' doesn't exist on the Hub

even though the user has access to the website NewT5/dummy_data since she/he is part of the org.

We need to improve the error message here similar to how @sgugger, @LysandreJik and @julien-c have done it for transformers IMO.

Steps to reproduce the bug

E.g. execute the following code to see the different error messages between transformes and datasets.

1. Transformers

from transformers import BertModel

BertModel.from_pretrained("NewT5/dummy_model")

The error message is clearer here - it gives:

OSError: patrickvonplaten/gpt2-xl is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'
If this is a private repository, make sure to pass a token having permission to this repo with `use_auth_token` or log in with `huggingface-cli login` and pass `use_auth_token=True`.

Let's maybe do the same for datasets? The PR was introduced to transformers here:
huggingface/transformers#15261

Expected results

Better error message

Actual results

Specify the actual results or traceback.

Environment info

• datasets version: 1.18.4.dev0
• Platform: Linux-5.15.15-76051515-generic-x86_64-with-glibc2.34
• Python version: 3.9.7
• PyArrow version: 6.0.1

[lhoestq]

We raise the error “ FileNotFoundError: can’t find the dataset” mainly to follow best practice in security (otherwise users could be able to guess what private repositories users/orgs may have)

We can indeed reformulate this and add the "If this is a private repository,..." part !

[mariosasko]

Resolved via #4536