What is the best way to save generate embeddings from BERT models?

Question

I want to find the best way to save, persist in disk the embedding from BERT models (transformers, sentence_transformers, ...), for the research tasks.