Datasets:

mmdjiji

/

bert-chinese-idioms

like 4

This breaks the dataset as train.json is 84,709 entries, whereas train_answers.csv 577,157 entries which results in

ERROR:datasets.packaged_modules.json.json:Failed to read file 'https://huggingface.co/datasets/mmdjiji/bert-chinese-idioms/resolve/d139909f9f053a68e2ef99acabe4f1d0d78c2ee1/train_answer.csv' with error <class 'pyarrow.lib.ArrowInvalid'>: JSON parse error: Invalid value. in row 0

Note: if anyone runs into this error you can simply cap the training examples or specify data_files = {"train": "train.json"}