Managing Container Best-Practices

I’m currently using docker with sshd and jupyterlab to develop our cuda code. One problem is that the machine freezes the machine on the critical path while the code runs. It affects the speed of other services running on the box. As the majority of the work is done on the GPU I don’t understand why the box would freeze.

So the question, Is there a standard way to manage heavy CUDA code using nvidia-docker ?