New techniques efficiently accelerate sparse tensors for massive AI models
New computational techniques, “HighLight” and “Tailors and Swiftiles,” could dramatically boost the speed and performance of high-performance computing applications like graph analytics or generative AI. The work, from MIT and NIVIDIA, aims to accelerate sparse tensors for AI models by introducing more efficient and flexible ways to take advantage of sparsity.
Joel Emer Taylor Swift Zi Yu Xue Yannan Nellie Wu Vivienne Sze International Symposium On Microarchitecture
Source: mit.edu