Timezone: »

Decentralized Training of Foundation Models in Heterogeneous Environments
Binhang Yuan · Yongjun He · Jared Davis · Tianyi Zhang · Tri Dao · Beidi Chen · Percy Liang · Christopher Ré · Ce Zhang

Wed Nov 30 09:00 AM -- 11:00 AM (PST) @ Hall J #225

Training foundation models, such as GPT-3 and PaLM, can be extremely expensive, often involving tens of thousands of GPUs running continuously for months. These models are typically trained in specialized clusters featuring fast, homogeneous interconnects and using carefully designed software systems that support both data parallelism and model/pipeline parallelism. Such dedicated clusters can be costly and difficult to obtain. Can we instead leverage the much greater amount of decentralized, heterogeneous, and lower-bandwidth interconnected compute? Previous works examining the heterogeneous, decentralized setting focus on relatively small models that can be trained in a purely data parallel manner. State-of-the-art schemes for model parallel foundation model training, such as Megatron and Deepspeed, only consider the homogeneous data center setting. In this paper, we present the first study of training large foundation models with model parallelism in a decentralized regime over a heterogeneous network. Our key technical contribution is a scheduling algorithm that allocates different computational “tasklets” in the training of foundation models to a group of decentralized GPU devices connected by a slow heterogeneous network. We provide a formal cost model and further propose an efficient evolutionary algorithm to find the optimal allocation strategy. We conduct extensive experiments that represent different scenarios for learning over geo-distributed devices simulated using real-world network measurements. In the most extreme case, across 8 different cities spanning 3 continents, our approach is 4.8× faster than prior state-of-the-art training systems.

Author Information

Binhang Yuan (ETH Zurich)
Yongjun He (ETHZ - ETH Zurich)
Jared Davis (Computer Science Department, Stanford University)
Tianyi Zhang (Stanford University)
Tri Dao (Stanford University)
Beidi Chen (Stanford University)
Percy Liang (Stanford University)
Christopher Ré (Stanford)
Ce Zhang (ETH Zurich)

More from the Same Authors