How to make load_dataset interpret "N/A" as a string and not convert to nan?

I am tryng to load a csv file using load_dataset. My file has 3 columns, 2 text sequences and and label. The text sequences are lines from technical documentation. Is there an easy way to tell load_dataset to interpret “N/A” as a string rather than convert it to nan?

Hi! You can pass keep_default_na=False to load_dataset to preserve nan values as string. We use pandas.read_csv to parse csv files in our loader, so more info on that param can be found here.

1 Like

That worked perfectly. Thank you!