The original model as GGUF Q8_0 (didn't tested smaller ones) have the issue to generate endless ' \n' after the prompt until the max context size is reached. That didn't happen here and works way better with chatml now.
· Sign up or log in to comment