Parallelizable Stack Long Short-Term Memory
arXiv:1904.03409
Abstract
Stack Long Short-Term Memory (StackLSTM) is useful for various applications such as parsing and string-to-tree neural machine translation, but it is also known to be notoriously difficult to parallelize for GPU training due to the fact that the computations are dependent on discrete operations. In this paper, we tackle this problem by utilizing state access patterns of StackLSTM to homogenize computations with regard to different discrete operations. Our parsing experiments show that the method scales up almost linearly with increasing batch size, and our parallelized PyTorch implementation trains significantly faster compared to the Dynet C++ implementation.
arXiv commentAccepted to NAACL 2019 Workshop on Structured Prediction for NLP
References in corpus (5)
- Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
- Transition-Based Dependency Parsing with Stack Long Short-Term Memory
- Revisiting Small Batch Training for Deep Neural Networks
- DyNet: The Dynamic Neural Network Toolkit
- Simple and Accurate Dependency Parsing Using Bidirectional LSTM Feature Representations