Escoin: Efficient Sparse Convolutional Neural Network Inference on GPUs

Abstract

Deep neural networks have achieved remarkable accuracy in many artificial intelligence applications, e.g. computer vision, at the cost of a large number of parameters and high computational complexity. Weight pruning can compress DNN models by removing redundant parameters in the networks, but it brings sparsity in the weight matrix, and therefore makes the computation inefficient on GPUs. Although pruning can remove more than 80% of the weights, it actually hurts inference performance (speed) when running models on GPUs. Two major problems cause this unsatisfactory performance on GPUs. First, lowering convolution onto matrix multiplication reduces data reuse opportunities and wastes memory bandwidth. Second, the sparsity brought by pruning makes the computation irregular, which leads to inefficiency when running on massively parallel GPUs. To overcome these two limitations, we propose Escort, an efficient sparse convolutional neural networks on GPUs. Instead of using the lowering method, we choose to compute the sparse convolutions directly. We then orchestrate the parallelism and locality for the direct sparse convolution kernel, and apply customized optimization techniques to further improve performance. Evaluation on NVIDIA GPUs show that Escort can improve sparse convolution speed by 2.63x and 3.07x, and inference speed by 1.43x and 1.69x, compared to CUBLAS and CUSPARSE respectively.

References in corpus (10)

• Caffe: Convolutional Architecture for Fast Feature Embedding
• Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
• Pruning Convolutional Neural Networks for Resource Efficient Inference
• Theano: new features and speed improvements
• cuDNN: Efficient Primitives for Deep Learning
• Exploring the Regularity of Sparse Structure in Convolutional Neural Networks
• Faster CNNs with Direct Sparse Convolutions and Guided Pruning
• Enabling Sparse Winograd Convolution by Native Pruning
• GARDENIA: A Domain-specific Benchmark Suite for Next-generation Accelerators
• Efficient Inferencing of Compressed Deep Neural Networks

Cited by in corpus (9)

• Only Train Once: A One-Shot Neural Network Training And Pruning Framework
• Low-Memory Neural Network Training: A Technical Report
• TIRAMISU: A Polyhedral Compiler for Dense and Sparse Deep Learning
• SparseRT: Accelerating Unstructured Sparsity on GPUs for Deep Learning Inference
• High Performance Convolution Using Sparsity and Patterns for Inference in Deep Convolutional Neural Networks
• Self-Supervised GAN Compression
• SparseDNN: Fast Sparse Deep Learning Inference on CPUs
• Accelerated CNN Training Through Gradient Approximation
• When deep learning models on GPU can be accelerated by taking advantage of unstructured sparsity