I am using NewsSentiment, a targeted sentiment analysis library for Python. It is (too) occasionally throwing the warning:
Be aware, overflowing tokens are not returned for the setting you have chosen, i.e. sequence pairs with the 'longest_first' truncation strategy. So the returned list will always be empty even if some tokens have been removed.
I have done some research and this appears to be issued by a huggingface transformer, and the library does have a Roberta dependency. However, I do not invoke Roberta in my code so I'm not sure how to suppress the warning. I have tried
warnings.filterwarnings("ignore", message="Be aware, overflowing tokens are not returned for the setting you have chosen, i.e. sequence pairs with the 'longest_first' truncation strategy. So the returned list will always be empty even if some toke")
warnings.filterwarnings("ignore", category=UserWarning, module='transformers')
But neither does the job. Anyone know how to do it?
Use this code snippet to get rid of this warning.