Token-Scaled Logit Distillation for Ternary Weight Generative Language Models

Minsoo Kim, Sihwa Lee, Janghwan Lee, Sukjin Hong, Du-Seong Chang, Wonyong Sung, Jungwook Choi

Keywords: Generative Language Model, Quantization, QAT, Knowledge Distillation, Causal Attention, Language Modeling

TL;DR: Investigate the unexplored challenges of Ternary QAT on GLMs and propose an innovative KD method to overcome them

Abstract: Generative Language Models (GLMs) have shown impressive performance in tasks such as text generation, understanding, and reasoning. However, the large model size poses challenges for practical deployment. To solve this problem, Quantization-Aware Training (QAT) has become increasingly popular. However, current QAT methods for generative models have resulted in a noticeable loss of accuracy. To counteract this issue, we propose a novel knowledge distillation method specifically designed for GLMs. Our method, called token-scaled logit distillation, prevents overfitting and provides superior learning from the teacher model and ground truth. This research marks the first evaluation of ternary weight quantization-aware training of large-scale GLMs with less than 1.0 degradation in perplexity and achieves enhanced accuracy in tasks like common-sense QA and arithmetic reasoning as well as natural language understanding. Our code is available at https://github.com/aiha-lab/TSLD.

Submission Number: 12651