加速大模型训练的双向交错流水线并行方法
吴厚明,陈岭

Bidirectional interleaved pipeline parallelism for accelerating large model training
Houming WU,Ling CHEN
图 7 流水线阶段数量扩展(流水线深度为4)
Fig.7 Scaling to larger pipeline stage numbers (pipeline depth is four)