NIPQ: Noise proxy-based Integrated Pseudo-Quantization
This addresses the problem of efficient neural network deployment for practitioners by improving quantization training, though it appears incremental as it builds on existing pseudoquantization frameworks.
The paper tackles the problem of unstable convergence and quality degradation in quantization-aware training by proposing NIPQ, a noise proxy-based integrated pseudoquantization method that unifies pseudoquantization for both activation and weight. The result shows that NIPQ outperforms existing quantization algorithms by a large margin in various vision and language applications.
Straight-through estimator (STE), which enables the gradient flow over the non-differentiable function via approximation, has been favored in studies related to quantization-aware training (QAT). However, STE incurs unstable convergence during QAT, resulting in notable quality degradation in low precision. Recently, pseudoquantization training has been proposed as an alternative approach to updating the learnable parameters using the pseudo-quantization noise instead of STE. In this study, we propose a novel noise proxy-based integrated pseudoquantization (NIPQ) that enables unified support of pseudoquantization for both activation and weight by integrating the idea of truncation on the pseudo-quantization framework. NIPQ updates all of the quantization parameters (e.g., bit-width and truncation boundary) as well as the network parameters via gradient descent without STE instability. According to our extensive experiments, NIPQ outperforms existing quantization algorithms in various vision and language applications by a large margin.