cnn_dailymail is broken in pinned version of datasets

[osanseviero]

Information

The problem arises in chapter:

• Introduction
• Text Classification
• Transformer Anatomy
• Multilingual Named Entity Recognition
• Text Generation
• Summarization
• Question Answering
• Making Transformers Efficient in Production
• Dealing with Few to No Labels
• Training Transformers from Scratch
• Future Directions

Describe the bug

As per huggingface/datasets#3830, trying to load the dataset fails. This change is still not in the latest release, but will likely need some update

To Reproduce

Steps to reproduce the behavior:

Just run

load_dataset("cnn_dailymail", '3.0.0')

For exact error message, see linked issue

[JingxinLee]

I met the same problem.

You could try

dataset = load_dataset('ccdv/cnn_dailymail', '3.0.0')

instead.

[osanseviero]

Doing

pip install git+https://github.com/huggingface/datasets#egg=datasets

and load_dataset("...", download_mode="force_redownload") should fix the issue, but it does require installing datasets from head

[lewtun]

Thanks for reporting the bug @osanseviero!

@lvwerra we should probably consider bumping datasets to the next release and checking that nothing breaks in the other chapters. One advantage of bumping to a later version is that we can then stream the audio dataset in the final chapter :)

[lvwerra]

I feel like we should avoid changing library versions globally as it requires a significant amount of work to make sure we don't break anything elsewhere. I'd rather fix it in the installation function for that chapter specifically.

[lewtun]

Sounds good @lvwerra! I'll take care of this :)

[albertvillanova]

@lewtun, this issue was fixed in datasets version 2.0.0.