DASH: Scheduling Deep Learning Workloads on Multi-Generational GPU-Accelerated Clusters | IEEE Conference Publication | IEEE Xplore