Loss Aware Post-training Quantization
This work addresses the accuracy drop in low-bit quantization for deploying large models on devices like mobile phones, representing an incremental improvement over existing post-training quantization techniques.
The paper tackled the problem of low-bit (e.g., INT4) neural network quantization for deployment on resource-constrained devices by analyzing the loss landscape, showing it becomes non-separable with steep curvature under aggressive quantization, and proposed a joint quantization method that significantly improves accuracy over current post-training methods.
Neural network quantization enables the deployment of large models on resource-constrained devices. Current post-training quantization methods fall short in terms of accuracy for INT4 (or lower) but provide reasonable accuracy for INT8 (or above). In this work, we study the effect of quantization on the structure of the loss landscape. Additionally, we show that the structure is flat and separable for mild quantization, enabling straightforward post-training quantization methods to achieve good results. We show that with more aggressive quantization, the loss landscape becomes highly non-separable with steep curvature, making the selection of quantization parameters more challenging. Armed with this understanding, we design a method that quantizes the layer parameters jointly, enabling significant accuracy improvement over current post-training quantization methods. Reference implementation is available at https://github.com/ynahshan/nn-quantization-pytorch/tree/master/lapq