State of the Art Neural NetworkCompression and Acceleration
How do you get reliable AI on the edge when it was made for the cloud? With millions or billions of parameters to process, today’s deep neural nets are too big for CPUs and MCUs to process in real time . Until now... Say hello to embedded AI that is fast, frugal and accurate. Context Adaptation, Quantization, Pruning and Batch Norm Folding formodels that are up to 330x smaller and 20x times faster.