@misc{wu2023understanding, title={Understanding INT4 Quantization for Transformer Models: Latency Speedup, Composability, and Failure Cases}, author={Xiaoxia Wu and Cheng Li and Reza Yazdani Aminabadi and Zhewei Yao and Yuxiong He}, year={2023}, eprint={2301.12017}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2301.12017} }