Training Deep Neural Networks with Joint Quantization and Pruning of Weights and Activations

Abstract

Quantization and pruning are core techniques used to reduce the inference costs of deep neural networks. State-of-the-art quantization techniques are currently applied to both the weights and activations; however, pruning is most often applied to only the weights of the network. In this work, we jointly apply novel uniform quantization and unstructured pruning methods to both the weights and activations of deep neural networks during training. Using our methods, we empirically evaluate the currently accepted prune-then-quantize paradigm across a wide range of computer vision tasks and observe a non-commutative nature when applied to both the weights and activations of deep neural networks. Informed by these observations, we articulate the non-commutativity hypothesis: for a given deep neural network being trained for a specific task, there exists an exact training schedule in which quantization and pruning can be introduced to optimize network performance. We identify that this optimal ordering not only exists, but also varies across discriminative and generative tasks. Using the optimal training schedule within our training framework, we demonstrate increased performance per memory footprint over existing solutions.

References in corpus (17)

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Empirical Evaluation of Rectified Activations in Convolutional Network
• Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations
• Quantizing deep convolutional networks for efficient inference: A whitepaper
• Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architectures
• To prune, or not to prune: exploring the efficacy of pruning for model compression
• The State of Sparsity in Deep Neural Networks
• Deep Learning Scaling is Predictable, Empirically
• Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation
• Sparse Networks from Scratch: Faster Training without Losing Performance
• Deep ReLU Networks Have Surprisingly Few Activation Patterns
• CondenseNet: An Efficient DenseNet using Learned Group Convolutions
• Bayesian Bits: Unifying Quantization and Pruning
• Towards the Limit of Network Quantization
• Understanding Straight-Through Estimator in Training Activation Quantized Neural Nets
• QGAN: Quantized Generative Adversarial Networks
• Joint Pruning & Quantization for Extremely Sparse Neural Networks

Cited by in corpus (2)

• Reducing Computational Complexity of Neural Networks in Optical Channel Equalization: From Concepts to Implementation
• DASS: Differentiable Architecture Search for Sparse neural networks