how to convert text to word embeddings using bert's pretrained model 'faster'?

Question

I\'m trying to get word embeddings for clinical data using microsoft/pubmedbert. I have 3.6 million text rows. Converting texts to vectors for 10k rows takes around 30 minut