Thank you!! I just loaded it a fits in memory as you said.

I am testing it now and seems quite fast giving the responses for a local model.