Overcoming Oscillations in Quantization-Aware Training
This addresses a critical issue in deploying efficient neural networks on resource-constrained devices, though it is incremental as it builds on existing QAT methods.
The paper tackles the problem of weight oscillations in quantization-aware training (QAT), which can cause significant accuracy degradation, especially in low-bit quantization of efficient networks like MobileNets and EfficientNets. The authors propose two novel QAT algorithms that achieve state-of-the-art accuracy for 3- and 4-bit quantization on ImageNet.
When training neural networks with simulated quantization, we observe that quantized weights can, rather unexpectedly, oscillate between two grid-points. The importance of this effect and its impact on quantization-aware training (QAT) are not well-understood or investigated in literature. In this paper, we delve deeper into the phenomenon of weight oscillations and show that it can lead to a significant accuracy degradation due to wrongly estimated batch-normalization statistics during inference and increased noise during training. These effects are particularly pronounced in low-bit ($\leq$ 4-bits) quantization of efficient networks with depth-wise separable layers, such as MobileNets and EfficientNets. In our analysis we investigate several previously proposed QAT algorithms and show that most of these are unable to overcome oscillations. Finally, we propose two novel QAT algorithms to overcome oscillations during training: oscillation dampening and iterative weight freezing. We demonstrate that our algorithms achieve state-of-the-art accuracy for low-bit (3 & 4 bits) weight and activation quantization of efficient architectures, such as MobileNetV2, MobileNetV3, and EfficentNet-lite on ImageNet. Our source code is available at {https://github.com/qualcomm-ai-research/oscillations-qat}.