RTP: Rethinking Tensor Parallelism with Memory Deduplication

November 02, 2023 ยท Entered Twilight ยท ๐Ÿ› arXiv.org

๐Ÿ’ค TWILIGHT: Eternal Rest
Repo abandoned since publication

Repo contents: LICENSE, README.md, benchmarks, docs, requirements.txt, rtp, rtp_mnist.py, setup.py, test.py, test_net.py, test_transformer.py, tests, tutorials

Authors Cheng Luo, Tianle Zhong, Geoffrey Fox arXiv ID 2311.01635 Category cs.DC: Distributed Computing Cross-listed cs.AI, cs.NI Citations 4 Venue arXiv.org Repository https://github.com/wdlctc/rtp โญ 11 Last Checked 3 months ago
Abstract
In the evolving landscape of neural network models, one prominent challenge stand out: the significant memory overheads associated with training expansive models. Addressing this challenge, this study delves deep into the Rotated Tensor Parallelism (RTP). RTP is an innovative approach that strategically focuses on memory deduplication in distributed training environments. It boasts of unique features like a customized communication primitive and the Flyweight Pattern initialization. Furthermore, RTP ensures a seamless overlap between partition computation and partition weight communication, optimizing the training process. Our empirical evaluations underscore RTP's efficiency, revealing that its memory consumption during distributed system training is remarkably close to the optimal - distributing the memory overhead of a single machine equitably among multiple machines. The experimental results demonstrate that RTP is capable of achieving comparable performance to Distributed Data Parallel while providing support for significantly larger models with near-linear scalability in terms of memory. Code of RTP is available at https://github.com/wdlctc/rtp.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

๐Ÿ“œ Similar Papers

In the same crypt โ€” Distributed Computing