Post-training 4-bit quantization of convolution networks for rapid-deployment

Abstract

Convolutional neural networks require significant memory bandwidth and storage for intermediate computations, apart from substantial computing resources. Neural network quantization has significant benefits in reducing the amount of intermediate results, but it often requires the full datasets and time-consuming fine tuning to recover the accuracy lost after quantization. This paper introduces the first practical 4-bit post training quantization approach: it does not involve training the quantized model (fine-tuning), nor it requires the availability of the full dataset. We target the quantization of both activations and weights and suggest three complementary methods for minimizing quantization error at the tensor level, two of whom obtain a closed-form analytical solution. Combining these methods, our approach achieves accuracy that is just a few percents less the state-of-the-art baseline across a wide range of convolutional models. The source code to replicate all experiments is available on GitHub: \url{https://github.com/submission2019/cnn-quantization}.

References in corpus (5)

• DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
• Quantizing deep convolutional networks for efficient inference: A whitepaper
• PACT: Parameterized Clipping Activation for Quantized Neural Networks
• Training and Inference with Integers in Deep Neural Networks
• Quantization for Rapid Deployment of Deep Neural Networks

Cited by in corpus (33)

• Applications and Techniques for Fast Machine Learning in Science
• Autoencoders on FPGAs for real-time, unsupervised new physics detection at 40 MHz at the Large Hadron Collider
• Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networks
• EasyQuant: Post-training Quantization via Scale Optimization
• Robust Quantization: One Model to Rule Them All
• Lightweight Compression of Intermediate Neural Network Features for Collaborative Intelligence
• DSConv: Efficient Convolution Operator
• Towards Unified INT8 Training for Convolutional Neural Network
• ZeroQ: A Novel Zero Shot Quantization Framework
• Post-Training Piecewise Linear Quantization for Deep Neural Networks
• Accelerated Sparse Neural Training: A Provable and Efficient Method to Find N:M Transposable Masks
• A Statistical Framework for Low-bitwidth Training of Deep Neural Networks
• A White Paper on Neural Network Quantization
• DRIVE: One-bit Distributed Mean Estimation
• Gradient ${\ell }_1$ Regularization for Quantization Robustness
• Diversifying Sample Generation for Accurate Data-Free Quantization
• Channel-wise Hessian Aware trace-Weighted Quantization of Neural Networks
• Full-Cycle Energy Consumption Benchmark for Low-Carbon Computer Vision
• Diverse Sample Generation: Pushing the Limit of Generative Data-free Quantization
• Neural gradients are near-lognormal: improved quantized and sparse training
• Qimera: Data-free Quantization with Synthetic Boundary Supporting Samples
• Optimal Gradient Quantization Condition for Communication-Efficient Distributed Training
• Dynamic Group Convolution for Accelerating Convolutional Neural Networks
• Post-Training Quantization for Vision Transformer
• Distribution-sensitive Information Retention for Accurate Binary Neural Network
• High Throughput Matrix-Matrix Multiplication between Asymmetric Bit-Width Operands
• Post-Training BatchNorm Recalibration
• SiMaN: Sign-to-Magnitude Network Binarization
• Memory Requirement Reduction of Deep Neural Networks Using Low-bit Quantization of Parameters
• HERO: Hessian-Enhanced Robust Optimization for Unifying and Improving Generalization and Quantization Performance
• Memory-Efficient CNN Accelerator Based on Interlayer Feature Map Compression
• Understanding and Overcoming the Challenges of Efficient Transformer Quantization
• HPTQ: Hardware-Friendly Post Training Quantization