1: \begin{abstract}%   <- trailing '%' for backward compatibility of .sty file

2: \texttt{gsplat} is an open-source library designed for training and developing Gaussian Splatting methods. It features a front-end with Python bindings compatible with the PyTorch library and a back-end with highly optimized CUDA kernels. \texttt{gsplat} offers numerous features that enhance the optimization of Gaussian Splatting models, which include optimization improvements for speed, memory, and convergence times. Experimental results demonstrate that \texttt{gsplat} achieves up to 10\% less training time and $4\times$ less memory than the original \cite{kerbl20233d} implementation. Utilized in several research projects, \texttt{gsplat} is actively maintained on GitHub. Source code is available at \url{https://github.com/nerfstudio-project/gsplat} under Apache License 2.0. We welcome contributions from the open-source community.

3: \end{abstract}

4: