Hi!
I've tested the model with both short and long sentences, it seems the outputs are random. Are you sure the checkpoint in repo contains trained weights ?
· Sign up or log in to comment
