1 paper
Yuan Yu, Martín Abadi, Paul Barham +12
Many recent machine learning models rely on fine-grained dynamic control flow for training and inference. In particular, models based on recurrent neural networks and on reinforcem…