@article{deepspeedulyssessystemoptimizationsfor, title = {DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models}, author = {Sam Ade Jacobs and Masahiro Tanaka and Chengming Zhang and Minjia Zhang and Shuaiwen Leon Song and Samyam Rajbhandari and Yuxiong He}, year = {2023}, eprint = {2309.14509}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2309.14509v2}, }