Datasets:

ivrit-ai

/

jpress-demo

like 0

Possibly the issue is with the image column format:

>>> import pandas as pd
>>> df= pd.read_parquet("hf://datasets/ivrit-ai/jpress-demo@~parquet/default/partial-train/0000.parquet")
df.shape
>>> df.shape
(55, 5)
>>> df.head(3)
                                               image                               uuid        date  segment                 label
0  {'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHD...  yehuda veyerusahlaim/1878.07.12/2  1878.07.12        2  yehuda veyerusahlaim
1  {'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHD...  yehuda veyerusahlaim/1878.07.12/5  1878.07.12        5  yehuda veyerusahlaim
2  {'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHD...  yehuda veyerusahlaim/1878.07.12/4  1878.07.12        4  yehuda veyerusahlaim

see also https://huggingface.co/datasets/UCSC-VLAA/HQ-Edit/discussions/2 and https://github.com/huggingface/dataset-viewer/issues/2215

cc @lhoestq

@severo thank you.

@lhoestq any thoughts on how to proceed with this?

I think it's the same issue as https://huggingface.co/datasets/arnaudstiegler/synthetic_us_passports_hard/discussions/1