Paper · arXiv 2311.02382
Ultra-Long Sequence Distributed Transformer
Xiao Wang, Isaac Lyngaas, Aristeidis Tsaris, Peng Chen, Sajal Dash, Mayanka Chandra Shekar, Tao Luo, Hong-Jun Yoon, Mohamed Wahib, John Gouley
The Long Short-Sequence Transformer (LSS Transformer) efficiently trains transformers on long sequences by segmenting them across GPUs with partial self-attention and novel communication techniques, achieving significant speed and memory improvements.
Abstract
Transformer models trained on long sequences often achieve higher accuracy than short sequences. Unfortunately, conventional transformers struggle with long sequence training due to the overwhelming computation and memory requirements. Existing methods for long sequence training offer limited speedup and memory reduction, and may compromise accuracy. This paper presents a novel and efficient distributed training method, the Long Short-Sequence Transformer (LSS Transformer), for training transformer with long sequences. It distributes a long sequence into segments among GPUs, with each GPU computing a partial self-attention for its segment. Then, it uses a fused communication and a novel double gradient averaging technique to avoid the need to aggregate partial self-attention and minimize communication overhead. We evaluated the performance between LSS Transformer and the state-of-the-art Nvidia sequence parallelism on a Wikipedia enwik8 dataset. Results show that our proposed method lead to 5.6x faster and 10.2x more memory-efficient implementation compared to state-of-the-art sequence parallelism on 144 Nvidia V100 GPUs. Moreover, our algorithm scales to an extreme sequence length of 50,112 at 3,456 GPUs, achieving 161% super-linear parallel efficiency and a throughput of 32 petaflops.