Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
Dask
License:

Format Error for Data Extraction or Loading

#3
by helloai0 - opened

Hi trying to use the data collection but encounter the format error when extracting the files by uncompress commands (e.g., zstd) or loading data (refer to another previous community comment)
Do you have a solution or a appropriate way to solve those? @monology

zstd -d val.jsonl.zst
image.png

It's working fine for me:

➜  zstd -d val.jsonl.zst
val.jsonl.zst       : 1042723385 bytes
➜  zstd --version
*** Zstandard CLI (64-bit) v1.5.5, by Yann Collet ***

Even with the datasets library (after pip install zstandard):

from datasets import load_dataset

data_files = "https://huggingface.co/datasets/monology/pile-uncopyrighted/resolve/main/train/00.jsonl.zst"
dataset = load_dataset("json", data_files=data_files, split="train")
dataset
Dataset({
    features: ['text', 'meta'],
    num_rows: 5899215
})

Now it works, thank you. @lorisrossi

helloai0 changed discussion status to closed

Sign up or log in to comment