Good models for few-shot multi-label text classification

I want to fine tune a pretrained model for multi label classification but only have a few hundred training examples. I know T5 can learn sequence to sequence generation pretty decently with only a few dozen examples. I’m wondering what are the go-to pretrained models for multilabel classification with limited training data?

I’ve had luck with autonlp/ autotrain for multi class classification but I need multilabel classifier. I could stack binary classifiers but that seems clumsy. Also, I’d like to know the base model in case I want to do some more pretraining with domain-particular text.