SQuAT: Sharpness- and Quantization-Aware Training for BERT
pith:KJ3VMQJ6open to challenge →
read the original abstract
Quantization is an effective technique to reduce memory footprint, inference latency, and power consumption of deep learning models. However, existing quantization methods suffer from accuracy degradation compared to full-precision (FP) models due to the errors introduced by coarse gradient estimation through non-differentiable quantization layers. The existence of sharp local minima in the loss landscapes of overparameterized models (e.g., Transformers) tends to aggravate such performance penalty in low-bit (2, 4 bits) settings. In this work, we propose sharpness- and quantization-aware training (SQuAT), which would encourage the model to converge to flatter minima while performing quantization-aware training. Our proposed method alternates training between sharpness objective and step-size objective, which could potentially let the model learn the most suitable parameter update magnitude to reach convergence near-flat minima. Extensive experiments show that our method can consistently outperform state-of-the-art quantized BERT models under 2, 3, and 4-bit settings on GLUE benchmarks by 1%, and can sometimes even outperform full precision (32-bit) models. Our experiments on empirical measurement of sharpness also suggest that our method would lead to flatter minima compared to other quantization methods.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
Zero-Shot Quantization via Weight-Space Arithmetic
A quantization vector derived from a donor model via weight-space arithmetic can be added to a receiver model to improve post-PTQ Top-1 accuracy by up to 60 points in 3-bit settings without receiver-side QAT or data.
-
Neural Network Quantization by Learning Low-Loss Subspaces
Learning quantization-aware linear paths in weight space yields a midpoint whose direct quantization matches quantization-aware training performance without using straight-through estimators.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.