I am running into a problem when creating a custom dataset where, upon tokenizing the data, the huggingface trainer seems to think my dataset has 0 samples. All features in my dataset shown in the attached image are arrays of length 100. Any idea why this might happen?
