Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networks
This work addresses the challenge of deploying efficient neural networks on resource-limited devices like mobile phones, representing an incremental improvement over existing quantization methods.
The paper tackles the problem of unstable training and performance degradation in low-bit neural network quantization by proposing Differentiable Soft Quantization (DSQ), which bridges full-precision and low-bit networks and achieves state-of-the-art results in experiments, with an efficient implementation showing up to 1.7x speedup on ARM devices.
Hardware-friendly network quantization (e.g., binary/uniform quantization) can efficiently accelerate the inference and meanwhile reduce memory consumption of the deep neural networks, which is crucial for model deployment on resource-limited devices like mobile phones. However, due to the discreteness of low-bit quantization, existing quantization methods often face the unstable training process and severe performance degradation. To address this problem, in this paper we propose Differentiable Soft Quantization (DSQ) to bridge the gap between the full-precision and low-bit networks. DSQ can automatically evolve during training to gradually approximate the standard quantization. Owing to its differentiable property, DSQ can help pursue the accurate gradients in backward propagation, and reduce the quantization loss in forward process with an appropriate clipping range. Extensive experiments over several popular network structures show that training low-bit neural networks with DSQ can consistently outperform state-of-the-art quantization methods. Besides, our first efficient implementation for deploying 2 to 4-bit DSQ on devices with ARM architecture achieves up to 1.7$\times$ speed up, compared with the open-source 8-bit high-performance inference framework NCNN. [31]