tensorflow.python.framework.errors_impl.InvalidArgumentError: [[{{node DecodeCSV}}]] [[IteratorGetNext]]

脚本报错：
tensorflow.python.framework.errors_impl.InvalidArgumentError: Expect 4 fields but have 3 in record 0
[[{{node DecodeCSV}}]]
[[IteratorGetNext]]
NLP中csv转TFRecord，train转正常， dev转失败，为0kb。
https://github.com/livingbody/EasyTransfer/tree/master/scripts/%E5%A4%A9%E6%B1%A0%E5%A4%A7%E8%B5%9B%E4%B8%93%E5%8C%BA

Inspect the scripts/天池大赛专区 directory linked in the issue and trace the dev CSV-to-TFRecord entry point. Reproduce the DecodeCSV failure and compare the dev record fields with the expected four-field format. Done means dev conversion completes without the field-count error and produces a non-empty TFRecord.

Written by the indexing model from the issue text.