Machine Learning Workload

Machine learning (ML) workloads encompass the computational demands of training and deploying ML models, a rapidly expanding field with significant energy and performance challenges. Current research focuses on optimizing various aspects, including benchmarking diverse model architectures (e.g., deep learning, transformer-based models) across different hardware platforms (CPUs, GPUs, specialized accelerators, and processing-in-memory systems), and developing efficient algorithms (like stochastic gradient descent) and schedulers to manage the dynamic nature of these workloads. These efforts aim to improve the efficiency, scalability, and performance of ML applications, impacting both scientific discovery and real-world deployments in areas like AIoT and high-performance computing.