New Techniques From MIT and NVIDIA Revolutionize Sparse Tensor Acceleration for AI

Complimentary approaches — "HighLight" and "Tailors and Swiftiles" — could boost the performance of demanding machine-le