Conglong Li, Ammar Ahmad Awan, Hanlin Tang, Samyam Rajbhandari, Yuxiong He. 1-bit LAMB: Communication Efficient Large-Scale Large-Batch Training with LAMB's Convergence Speed. In 29th IEEE International Conference on High Performance Computing, Data, and Analytics, HiPC 2022, Bengaluru, India, December 18-21, 2022. pages 272-281, IEEE, 2022. [doi]
Abstract is missing.