Datasets:
Dataset scripts are no longer supported, but found Spatial457.py
Dear Authors,
I hope my email finds you well.
I want to use your benchmark in research but found the error: RuntimeError: Dataset scripts are no longer supported, but found Spatial457.py. Maybe it is a good way to use .parquet to store image-text dataset? So you don't need to separate images and texts like the current version.
Thank you for your time!
Kind regards,
Hi @kdeng03 . Thanks for sharing the bug. Could you share your code for me to reproduce?
Hi,
Just the script from README.md:
from datasets import load_dataset
# Load dataset from Hugging Face Hub
dataset = load_dataset(
"RyanWW/Spatial457",
name="L5_6d_spatial",
split="validation",
trust_remote_code=True # Required for custom loading script
)
Hi @kdeng03 ,
Thanks for confirming! Could you share your datasets version? import datasets print(datasets.__version__)
We tested with datasets==3.5.0 and trust_remote_code=True, it loads fine on our end. If you're on a newer version, try: pip install "datasets==3.5.0"
We will consider migrating to Parquet format in a future update to improve compatibility. Thanks for the suggestion!
Sorry for the inconvenience.
Hi @kdeng03 ,
Thanks for confirming! Could you share your datasets version?
import datasets print(datasets.__version__)We tested with
datasets==3.5.0andtrust_remote_code=True, it loads fine on our end. If you're on a newer version, try:pip install "datasets==3.5.0"We will consider migrating to Parquet format in a future update to improve compatibility. Thanks for the suggestion!
Sorry for the inconvenience.
Hi,
My version is 4.8.5. I also got this warning:
`trust_remote_code` is not supported anymore.
Please check that the Hugging Face dataset 'RyanWW/Spatial457' 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.
So probably hf removed trust_remote_code in newer version of datasets.
Thank you for your time again!
Hi @kdeng03 ,
You're right, newer versions of datasets are no longer compatible with our loading script. For now, please downgrade to datasets==3.5.0 as a workaround: pip install "datasets==3.5.0"
We will update the dataset to a standard Parquet format in a future release.
Thanks for your patience!