The maximum number of GPU accelerators in the worker host.
The minimum number of GPU accelerators in the worker host.