OOM on larger models

Hi Authors,

Thank you for sharing your code! However, I run into an OOM error on larger models (such as Llama-7b or Vicuna-7b). I am using 80G A-100 GPUs. Could you share your configurations on these models? Thank you!