Similar to how there are already metrics for non-GPU environments that enable users to see their utilization and make optimization decisions, add support for GPU environments as well.