PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel
Authors:
Yanli Zhao,
Andrew Gu,
Rohan Varma,
Liang Luo,
Chien-Chin Huang,
Min Xu,
Less Wright,
Hamid Shojanazeri,
Myle Ott,
Sam Shleifer,
Alban Desmaison,
Can Balioglu,
Pritam Damania,
Bernard Nguyen,
Geeta Chauhan,
Yuchen Hao,
Ajit Mathews,
Shen Li
Abstract:
It is widely acknowledged that large models have the potential to deliver superior performance across a broad range of domains. Despite the remarkable progress made in the field of machine learning systems research, which has enabled the development and exploration of large models, such abilities remain confined to a small group of advanced users and industry leaders, resulting in an implicit tech…
▽ More
It is widely acknowledged that large models have the potential to deliver superior performance across a broad range of domains. Despite the remarkable progress made in the field of machine learning systems research, which has enabled the development and exploration of large models, such abilities remain confined to a small group of advanced users and industry leaders, resulting in an implicit technical barrier for the wider community to access and leverage these technologies. In this paper, we introduce PyTorch Fully Sharded Data Parallel (FSDP) as an industry-grade solution for large model training. FSDP has been closely co-designed with several key PyTorch core components including Tensor implementation, dispatcher system, and CUDA memory caching allocator, to provide non-intrusive user experiences and high training efficiency. Additionally, FSDP natively incorporates a range of techniques and settings to optimize resource utilization across a variety of hardware configurations. The experimental results demonstrate that FSDP is capable of achieving comparable performance to Distributed Data Parallel while providing support for significantly larger models with near-linear scalability in terms of TFLOPS.
△ Less
Submitted 12 September, 2023; v1 submitted 21 April, 2023;
originally announced April 2023.
