Stars
C++ implementation of the Python Numpy library
FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.
A technical report on convolution arithmetic in the context of deep learning
VSCode extension for code suggestion
A simple tool that can generate TensorRT plugin code quickly.
AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.
Concurrent CPU-GPU Programming using Task Models
heterogeneity-aware-lowering-and-optimization
Automatic Mapping Generation, Verification, and Exploration for ISA-based Spatial Accelerators
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
Deep Learning tools and applications for NVIDIA AGX platforms.
[ARCHIVED] The C++ parallel algorithms library. See https://github.com/NVIDIA/cccl
Online CUDA Occupancy Calculator
User space software for Intel(R) Resource Director Technology
PyTorch/TorchScript/FX compiler for NVIDIA GPUs using TensorRT
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
This patch removes restriction on maximum number of simultaneous NVENC video encoding sessions imposed by Nvidia to consumer-grade GPUs.
Autoware - the world's leading open-source software project for autonomous driving
ChatterBot is a machine learning, conversational dialog engine for creating chat bots
My fork of Alex Krizhevsky's cuda-convnet from 2013 where I added dropout, among other features.
Protothread.h, a C++ port of Adam Dunkels' protothreads library
The C++ Core Guidelines are a set of tried-and-true guidelines, rules, and best practices about coding in C++
A basic implementation of the D* lite algorithm
Contains high quality implementations of Deep Reinforcement Learning algorithms written in PyTorch
Some basic examples of playing with RL