Reproduce “ByteSplat: Efficient Distributed 3D Gaussian Splatting Training via Intra- and Inter-GPU communication reduction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shuo Wu, He Zhu, Han Zhao, Xiaohui Zhang, Yaqian Zhao, Hui Wei, Ruyang Li, Hongzhi Shi and 4 more
ByteSplat reduces data movement in distributed 3D Gaussian Splatting training by keeping reusable rendering data on the GPU, pruning with per-tile storage limits in mind, and skipping exactly zero partial gradients. Across six datasets on eight A6000 GPUs, it reduces per-iteration latency by an average 3.4×, reaching 6.1× on UrbanScene3D versus Grendel-GS, with small measured image-quality changes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.