VRay has an environment variable (VRAY_GPU_PLATFORMS) that you can set on each render slave. VRay will look to this environment variable on a render slave to determine which GPUs are to be used when rendering on that slave. With such a variable, each render slave can have a different GPU configuration from other render slaves and the render client submitting the request. When preparing a render from the render client, you would set the render type to GPU and identify only the GPUs to use on the render client. VRay would then rely on the environment variable on each render slave to determine the exact GPU resources to use on that slave.
The documentation for Prorender Maya indicates that, when preparing a render from the render client, you need to select GPU and identify the GPUs resources to use. It does not indicate whether or not there is an environment variable to set on each render slave to identify the GPU resources to use on those slaves.
Given this context, my question really is...
How do I render on multiple render slaves that have different GPUs than the render client?
A generic response is to use Batch rendering. However, it does not explain how to do this with multiple machines that have different GPU configurations.
Any help would be appreciated.