How to get features from RoBERTa model classifier

I am using RoBERTa pretrained model go classify text.
model = RobertaForSequenceClassification.from_pretrained(‘roberta-base’)
outputs = model(**source_batch)

I want to know which parameter returns the features of the input ?

I guess the RobertaConfig defines the number of input features as well as other hyperparameters.
In any case, you might also want to post this question in the HuggingFace discussion board, since the code owners would see it there, too. :slight_smile: