Code and data for paper "Deep Painterly Harmonization": https://arxiv.org/abs/1804.03189

FlashInfer: Kernel Library for LLM Serving

Fast parallel CTC.

Squeeze-and-Excitation Networks

Fully Convolutional Instance-aware Semantic Segmentation

Efficient GPU kernels for block-sparse matrix multiplication and convolution

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing …

Training materials associated with NVIDIA's CUDA Training Series (www.olcf.ornl.gov/cuda-training-series/)

CUDA Kernel Benchmarking Library

Reference implementation of real-time autoregressive wavenet inference

An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).

Parrot is a C++ library for fused array operations using CUDA/Thrust. It provides efficient GPU-accelerated operations with lazy evaluation semantics, allowing for chaining of operations without un…

PyTorch bindings for CUTLASS grouped GEMM.

WholeGraph - large scale Graph Neural Networks

Optimized Parallel Tiled Approach to perform Matrix Multiplication by taking advantage of the lower latency, higher bandwidth shared memory within GPU thread blocks.