What is the fast way to get tensorflow universal sentence embeddings on large corpus?

Question

I have a corpus of 100k rows with on an average 20 sentences in each row, stored in pandas column. What is the fastest way to get tensorflow universal sentence embedding for eac