batchsize 16 on one 2080ti cause OOM exception

[fyture]

I encounter with a problem when setting batchsize 16 on one 2080ti 11G causes OOM. So I have to put 8 on two GPUs respectively. But the paper says the experiment train only on one 1080ti, So I am so confused and don't know what I am missing. And also, while training, my GPU memory are full. But the use of GPU is 0%. Could you share me with some advices? Thank you very much.