How to use word embeddings from bert/use in lstm model for NER problem?

Question

I have 50 documents and ner tags of each words in every sentence. I have created bert embeddings for 50 documents and tried to pass that to lstm embeddings.