LLAMA-2 conversation generated responses always empty

The solution is to give max_length:

llama = pipeline(
    "conversational",
    model="meta-llama/Llama-2-7b-chat-hf",
    max_length=2000
)
1 Like