I understand that we can set virtual resources to workers and specify task resources.
Assume, that we have a local cluster with 2 Nvidia GPUs and created two workers each