On-the-fly Operation Batching in Dynamic Computation Graphs

Abstract

Dynamic neural network toolkits such as PyTorch, DyNet, and Chainer offer more flexibility for implementing models that cope with data of varying dimensions and structure, relative to toolkits that operate on statically declared computations (e.g., TensorFlow, CNTK, and Theano). However, existing toolkits - both static and dynamic - require that the developer organize the computations into the batches necessary for exploiting high-performance algorithms and hardware. This batching task is generally difficult, but it becomes a major hurdle as architectures become complex. In this paper, we present an algorithm, and its implementation in the DyNet toolkit, for automatically batching operations. Developers simply write minibatch computations as aggregations of single instance computations, and the batching algorithm seamlessly executes them, on the fly, using computationally efficient batched operations. On a variety of tasks, we obtain throughput similar to that obtained with manual batches, as well as comparable speedups over single-instance learning on architectures that are impractical to batch manually.

References in corpus (7)

• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• Transition-Based Dependency Parsing with Stack Long Short-Term Memory
• DyNet: The Dynamic Neural Network Toolkit
• Deep Learning with Dynamic Computation Graphs
• Learning to Compose Words into Sentences with Reinforcement Learning
• Caffe con Troll: Shallow Ideas to Speed Up Deep Learning
• Training with Exploration Improves a Greedy Stack-LSTM Parser

Cited by in corpus (3)

• TensorFlow-Serving: Flexible, High-Performance ML Serving
• Hierarchical Character Embeddings: Learning Phonological and Semantic Representations in Languages of Logographic Origin using Recursive Neural Networks
• Cavs: A Vertex-centric Programming Interface for Dynamic Neural Networks