karpathy / llm.c
LLM training in simple, raw C/CUDA
See what the GitHub community is most excited about today.
LLM training in simple, raw C/CUDA
FlashInfer: Kernel Library for LLM Serving
DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling
DeepEP: an efficient expert-parallel communication library
CUDA accelerated rasterization of gaussian splatting
NCCL Tests
Tile primitives for speedy kernels
CUDA Kernel Benchmarking Library
[ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without lossing end-to-end metrics across language, image, and video models.
CUDA Library Samples
GPU accelerated decision optimization
Fast CUDA matrix multiplication from scratch
Causal depthwise conv1d in CUDA, with a PyTorch interface
cuVS - a library for vector search and clustering on the GPU
PyTorch bindings for CUTLASS grouped GEMM.
Instant neural graphics primitives: lightning fast NeRF and more
[ARCHIVED] Cooperative primitives for CUDA C++. See https://github.com/NVIDIA/cccl