Text this: Architecture of multiple processing units for accelerating deep neural network inference /