Embeddings from llama2

I have the same situation as mentioned by Saaira.
Does anyone have any solutions or explanation for this?
Thanks!