Effective Approaches to Batch Parallelization for Dynamic Neural Network Architectures

Re-assign community