Machine learning models, especially deep neural networks are tend to be computationally heavy. To reduce these costs it is possible to optimize the performance of the networks. A deep learning compiler can provide over 10x acceleration.