Add truncated llama style model init via reset parameters()#54