It may be possible to enable accelerated workflows for Apple GPU users by adding CoreML to the execution providers list.
The tricky part of doing this will be detecting when this execution provider should or should not be prepended to the executors list.