@article{m610tasharingdelinkingparadigmfor, title = {M6-10T: A Sharing-Delinking Paradigm for Efficient Multi-Trillion Parameter Pretraining}, author = {Junyang Lin and An Yang and Jinze Bai and Chang Zhou and Le Jiang and Xianyan Jia and Ang Wang and Jie Zhang and Yong Li and Wei Lin and Jingren Zhou and Hongxia Yang}, year = {2021}, eprint = {2110.03888}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2110.03888v3}, }