LLM training in simple, raw C/CUDA

DeepEP: an efficient expert-parallel communication library

[MICRO'23, MLSys'22] TorchSparse: Efficient Training and Inference Framework for Sparse Convolution on GPUs.

Differentiable Iso-Surface Extraction Package (DISO)

Code for SpotLessSplats: Ignoring Distractors in 3D Gaussian Splatting built on gsplat codebase.

An extension to official diff-gaussian-rasterization with forward and backward depth pass

The real-time viewer and offline renderer for the HPG'24 paper "Frustum Volume Caching for Accelerated NeRF Rendering"

Utilizing the texture and depth priors to initialize the Gaussian points