Hi
I can use 4 GPUs and I want to split the deep learning model in half and run it using the CUDA stream.
Then am I right to use the MultistreamGuard?
Is it right to change the GPU number and stream number at the same time using the MutistreamGuard?