Looking at the code in roberta modeling it appears that
RobertaForSequenceClassification
can only use the “< s>” token as “pooler” is this true or am I looking at the wrong code?
Thanks to anybody who can answer!
Looking at the code in roberta modeling it appears that
RobertaForSequenceClassification
can only use the “< s>” token as “pooler” is this true or am I looking at the wrong code?
Thanks to anybody who can answer!