Using second gpu?

Both TensorFlow and PyTorch has a method to limit which GPUs are visible.

For example, os.environ["CUDA_VISIBLE_DEVICES"]="1"