LLM training in simple, raw C/CUDA

DeepEP: an efficient expert-parallel communication library

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

Tile primitives for speedy kernels

CUDA Library Samples

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

Automatically exported from code.google.com/p/cuda-convnet2