load dataset failed
#3
by
ustc-zhangzm
- opened
Hi, I try to load the datasets with the provided script
from datasets import load_dataset
# Load AetherCode v1 2024 (Jan 2024 - Dec 2024)
ds = load_dataset("m-a-p/AetherCode", "v1_2024")
# Load AetherCode v1 2025 (Jan 2025 - May 2025)
ds = load_dataset("m-a-p/AetherCode", "v1_2025")
but failed with ArrowNotImplementedError: Nested data conversions not implemented for chunked array outputs.
It seems that this is caused when loading v1_2024/test-00025-of-00041.parquet, which is the largest parquet file.
Same problem emerges when loading the ByteDance-Seed/Code-Contests-Plus 3x split, which shares the same first author with this repo.
I wonder is there any extra config to successfully load these two datasets?