ValueError: The train dataset must contain a 'tokens' column.

[RupaliKapatel]

Spanmarker version : 1.7.0

{'text': 'Work closely with business leads and product owners.',
'entities': [{'end': 51, 'label': 'compound_task', 'start': 0},
{'end': 87, 'label': 'task', 'start': 55},
{'end': 123, 'label': 'task', 'start': 92}]}

It giving error : ValueError: The train dataset must contain a 'tokens' column.

I want to train for nested spans .... dont want to use BIO or other token level tagging.