Skip to content

Conversation

lhoestq
Copy link
Member

@lhoestq lhoestq commented Jun 4, 2025

TODO:

  • remplace fixtures based on script with no-script fixtures
  • windaube

@HuggingFaceDocBuilderDev

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

@lhoestq lhoestq marked this pull request as ready for review June 9, 2025 16:45
@lhoestq lhoestq merged commit 784607d into main Jun 9, 2025
9 of 15 checks passed
@lhoestq lhoestq deleted the remove-scripts-altogether branch June 9, 2025 16:45
@jp1924
Copy link
Contributor

jp1924 commented Jul 14, 2025

Hi @lhoestq,
I wanted to ask
are you planning to stop supporting dataset builds using GeneratorBasedBuilder?

If so, could you share the reason why?

@lhoestq
Copy link
Member Author

lhoestq commented Jul 15, 2025

We stopped supporting dataset scripts altogether, whether they are based on GeneratorBasedBuilder or any other builder. This means you can't load_dataset() a dataset script anymore. We did this mostly for security reasons which is blocking for many users and also impossible to build upon (e.g. the for the Dataset Viewer on HF)

@jp1924
Copy link
Contributor

jp1924 commented Jul 16, 2025

Ah, so only the trust_remote_code feature of load_dataset is deprecated, and

from datasets import load_dataset_builder
  
builder = load_dataset_builder('cornell-movie-review-data/rotten_tomatoes')  
builder.download_and_prepare()  

we can still load data using load_dataset_builder and download_and_prepare, right?
That's a relief. I thought the removal of trust_remote_code in load_dataset meant GeneratorBasedBuilder was being deprecated too, haha.
Got it, thanks for the clarification!

@autosquid
Copy link

Can you give an example on how to upgrade from using trust_remote_code? I used to load_dataset from a script generating my training data in a streaming way.

@autosquid
Copy link

autosquid commented Aug 4, 2025

For guys who dislike this change +1

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants