NOVA: NoC-based Vector Unit for Mapping Attention Layers on a CNN Accelerator

arXiv:2405.04206v1 Announce Type: cross
Abstract: Attention mechanisms are becoming increasingly popular, being used in neural network models in multiple domains such as natural language processing (NLP) and vision applications, especially at the edge. However, attention layers are difficult to map onto existing neuro accelerators since they have a much higher density of non-linear operations, which lead to inefficient utilization of today's vector units. This work introduces NOVA, a NoC-based Vector Unit that can perform non-linear operations within the NoC of …

abstract accelerator accelerators applications arxiv attention attention mechanisms cnn cs.ai cs.ar cs.lg domains edge however language language processing map mapping multiple natural natural language natural language processing network neural network neuro nlp nova popular processing the edge type vector vision