Evaluate the need for a pegasus-specific profile variable (e.g. pegasus_gpus) that requests N number of gpu resources.
This variable can be used as a flag to enable GPU exposure to containers.

An alternative is the currently present profile "request_gpus"