Towards Scalability and Performance: Framework for Heterogeneous Cluster Integration in Deep Learning Accelerators

This paper presents a streamlined approach to address the challenges of integrating heterogeneous clusters in deep learning(DL) accelerators. As the demand for scalable and high-performance computing in DL continues to grow, the proposed framework offers an easy implementation solution, ensuring efficiency without compromise. The method details core components and design principles, guiding the inclusion of heterogeneous clusters step by step. Extensive experiments show the framework's adaptability, handling diverse workloads and cluster architectures. Our exploration integrates various dataflows, dynamically selected by NN during runtime for optimal performance. Results indicate improved scalability and robustness compared to existing methods. This research provides a valuable resource for practitioners seeking efficient and accessible solutions for harnessing the power of heterogeneous clusters in DL acceleration, with implications for advancements in AI and computational efficiency.