Data-Free Quantization Through Weight Equalization and Bias Correction
This addresses the challenge of efficient inference on hardware for computer vision practitioners, offering a data-free solution that reduces engineering time and performance loss, though it is incremental as it builds on existing quantization techniques.
The paper tackles the problem of quantizing deep neural networks to 8-bit without requiring fine-tuning or hyperparameter selection, achieving near-original model performance and state-of-the-art results on architectures like MobileNet, with extensions to tasks such as semantic segmentation and object detection.
We introduce a data-free quantization method for deep neural networks that does not require fine-tuning or hyperparameter selection. It achieves near-original model performance on common computer vision architectures and tasks. 8-bit fixed-point quantization is essential for efficient inference on modern deep learning hardware. However, quantizing models to run in 8-bit is a non-trivial task, frequently leading to either significant performance reduction or engineering time spent on training a network to be amenable to quantization. Our approach relies on equalizing the weight ranges in the network by making use of a scale-equivariance property of activation functions. In addition the method corrects biases in the error that are introduced during quantization. This improves quantization accuracy performance, and can be applied to many common computer vision architectures with a straight forward API call. For common architectures, such as the MobileNet family, we achieve state-of-the-art quantized model performance. We further show that the method also extends to other computer vision architectures and tasks such as semantic segmentation and object detection.