Datasets 4.0.0 RuntimeError: Dataset scripts are no longer supported, but found GAIA.py
In the latest version of datasets 4.0.0, they have removed support of using a loading script, which this dataset has at GAIA.py.
At the current version we're unable to use the dataset.
Full error:
Please check that the Hugging Face dataset '.../gaia_dataset/GAIA' isn't based on a loading script and remove trust_remote_code.
If the dataset is based on a loading script, please ask the dataset author to remove it and convert it to a standard format like Parquet.
Required fix:
Change dataset to no longer rely on the loading script, as the error suggests
Hi! Yes, I have not yet had the time to change this. and I don't have the bandwidth at the moment.
If you want to open a PR to update the dataset, that would be delightful
@NelsonGc
@clefourrier
I’ve opened a PR (discussion #42) with a Parquet-converted version of the GAIA dataset.
As far as I know, there’s no way to have a single column in datasets that automatically loads mixed file types (PDF, PNG, CSV, etc.)—auto-decoding only works for typed media columns like image, audio, or pdf.
Given that GAIA’s attachments are heterogeneous, I suggest we keep the file_path column so users can still easily fetch the original files.