Accurate Neural Training with 4-bit Matrix Multiplications at Standard Formats

TL;DR

This paper introduces a novel 4-bit quantization method for neural network training that combines unbiased and logarithmic quantization, enabling efficient training with minimal accuracy loss.

Contribution

The authors propose the logarithmic unbiased quantization (LUQ) method, which effectively quantizes both forward and backward passes to 4-bit, improving training accuracy and efficiency.

Findings

Achieved 1.1% degradation on ResNet50 with 4-bit training.

Reduced degradation to 0.32% after fine-tuning.

State-of-the-art results in 4-bit neural network training.

Abstract

Quantization of the weights and activations is one of the main methods to reduce the computational footprint of Deep Neural Networks (DNNs) training. Current methods enable 4-bit quantization of the forward phase. However, this constitutes only a third of the training process. Reducing the computational footprint of the entire training process requires the quantization of the neural gradients, i.e., the loss gradients with respect to the outputs of intermediate neural layers. Previous works separately showed that accurate 4-bit quantization of the neural gradients needs to (1) be unbiased and (2) have a log scale. However, no previous work aimed to combine both ideas, as we do in this work. Specifically, we examine the importance of having unbiased quantization in quantized neural network training, where to maintain it, and how to combine it with logarithmic quantization. Based on this, we suggest a $\textit{logarithmic unbiased quantization}$ (LUQ) method to quantize both the forward and backward phases to 4-bit, achieving state-of-the-art results in 4-bit training without the overhead. For example, in ResNet50 on ImageNet, we achieved a degradation of 1.1%. We further improve this to a degradation of only 0.32% after three epochs of high precision fine-tuning, combined with a variance reduction method -- where both these methods add overhead comparable to previously suggested methods.