A Linear Algebraic Approach to Model Parallelism in Deep Learning
This addresses the challenge of memory and processing limitations in large-scale deep learning for researchers and practitioners, though it is incremental as it builds on existing distributed tools.
The paper tackles the problem of training large deep neural networks across distributed computing clusters by proposing a linear-algebraic approach to model parallelism, enabling parallel distribution of any tensor and manually deriving adjoint operators for gradient-based training, with implementation demonstrated using DistDL on PyTorch and MPI.
Training deep neural networks (DNNs) in large-cluster computing environments is increasingly necessary, as networks grow in size and complexity. Local memory and processing limitations require robust data and model parallelism for crossing compute node boundaries. We propose a linear-algebraic approach to model parallelism in deep learning, which allows parallel distribution of any tensor in the DNN. Rather than rely on automatic differentiation tools, which do not universally support distributed memory parallelism models, we show that parallel data movement operations, e.g., broadcast, sum-reduce, and halo exchange, are linear operators, and by defining the relevant spaces and inner products, we manually develop the adjoint, or backward, operators required for gradient-based training of DNNs. We build distributed DNN layers using these parallel primitives, composed with sequential layer implementations, and demonstrate their application by building and training a distributed DNN using DistDL, a PyTorch and MPI-based distributed deep learning toolkit.