GPU

GPU = graphical processing unit, which is a dedicated type of computer hardware that executes many instructions in parallel, operating on parallel data streams. This results in significantly faster processing for parallel neural network models relative to a standard serial CPU (central processing unit). The ability to leverage GPU hardware for neural network models, popularized by the Krizhevsky et al. (2012) AlexNet paper, is critical for making it possible to train and run the current large language models.