PRT: An Efficient Pipeline Reuse Technology for Large Models Training

The rapid evolution of large models and the widespread application of extensive datasets have made the cost of training increasingly prohibitive. While pipeline model parallelism makes it possible to train large models, existing pipeline techniques find it difficult to reduce bubble time due to thei...

Full description

Bibliographic Details

Published in:	Proceedings / IEEE International Conference on Cluster Computing pp. 1 - 11
Main Authors:	Ji, Zeyu, Zhai, Banghao, Zhang, Zhonghao, Chu, Qi, Liu, Bin
Format:	Conference Proceeding
Language:	English
Published:	IEEE 02.09.2025
Subjects:	Bidirectional control Computational modeling Distributed training Encoding Memory management Parallel processing Pipeline Model Parallelism Pipeline Reuse Technology Pipelines Resource management Servers Throughput Training
ISSN:	2168-9253
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!