j-hartmann

/

sentiment-roberta-large-english-3-classes

like 18

I am trying to get sentiment scores for news articles. But I am facing issue of input length even after proprocessing.
Is there a good way to shorten the length without losing relevant information?

Error:
Input is too long, try to truncate or use a paramater to handle this: The expanded size of the tensor (4099) must match the existing size (514) at non-singleton dimension 1. Target sizes: [1, 4099]. Tensor sizes: [1, 514]