1 paper
Ranganath R. Selagamsetty, Matthew Poremba, Bradford M. Beckmann +2
As distributed AI workloads grow in scale, multi-GPU systems have become essential for training large models. Although techniques like kernel fusion and overlapping communication w…