Open Source at NVIDIA

Engaging with open-source communities accelerates innovation, making it easier for developers to collaborate and build. NVIDIA contributes to important open-source projects—including Docker, JAX, Kubernetes, Linux kernel, PyTorch, TensorFlow, and Universal Scene Description (USD)—and leads innovative open-source projects across many domains. NVIDIA also actively contributes to and collaborates with open-standards bodies worldwide.

Explore the innovative open-source projects below to learn more about accelerating your applications.

All

Apache

Apache Spark

GPU-accelerated Apache Spark for data analytics, machine learning, and deep learning pipelines

NVIDIA

Apex

A PyTorch extension: Tools for easy mixed precision and distributed training in PyTorch

NVIDIA

AmgX

Distributed multi-grid linear solver library on GPU

NVIDIA

Clara Holoscan SDK

A hybrid computing platform for medical devices, both hardware and software

NVIDIA

Cloud Native Stack

A reference architecture for managing GPU workloads with Kubernetes

NVIDIA

Collective Communication Library (NCCL)

Optimized primitives for collective multi-GPU communication

NVIDIA

Container Toolkit

Allows users to build and run GPU accelerated containers.

NVIDIA

CUB

A library of collective primitives and utilities

NVIDIA

CUDA Templates for Linear Algebra Subroutines (CUTLASS)

A collection of templates for high-performance general matrix multiplication (GEMM) computations within CUDA kernels

NVIDIA

CuPy

An open-source array library for GPU-accelerated computing with Python and CUDA

NVIDIA

Data Loading Library (DALI)

Data pre-processing in deep learning applications

NVIDIA

Deep Learning Inference Accelerator (NVDLA)

Open-source accelerator for deep learning inference

NVIDIA

Falcor

A real-time rendering framework

NVIDIA

FasterTransformer

Highly optimized transformer-based encoder and decoder component for GPT and BERT models

NVIDIA

Federated Learning Active Runtime Environment (FLARE)

An open-source, extensible SDK that allows researchers and data scientists to adapt existing machine learning and deep learning workflow to a privacy-preserving federated paradigm

Flang

Flang

A Fortran compiler targeting low-level virtual machine (LLVM)

NVIDIA

GPU Operator

Automates the lifecycle management of the software required to expose GPUs on Kubernetes.

NVIDIA

K8s Device Plug-in

An NVIDIA plug-in that enables GPU support in Kubernetes

NVIDIA

Kaolin Library

PyTorch-based library with modular optimized functionality for 3D deep learning research

NVIDIA

libcu++

C++ standard library for an entire CPU+GPU system supported by CUDA

NVIDIA

Material Definition Language (MDL)

Physically accurate material standard and SDK

NVIDIA

Megatron-LM

Ongoing research for training transformer models at scale

NVIDIA

Merlin

Open-source framework for building high-performing recommender systems at scale

NVIDIA

Milano

Milano is a tool for automating hyper-parameters search for your models on a backend of your choice

NVIDIA

Modulus

A neural network framework for developing Physics-ML models

NVIDIA

MONAI

PyTorch-based framework for deep learning in healthcare imaging

NVIDIA

Morpheus

Open-source framework that enables cybersecurity developers to create optimized AI pipelines for filtering, processing, and classifying large volumes of real-time data

NVIDIA

NeMo Guardrails

A toolkit for easily developing trustworthy, safe, and secure LLM conversational systems.

NVIDIA

Network Operator

Automates the lifecycle management of the software for accelerated networking on Kubernetes.

NVIDIA

nvcomp

High-performance GPU data compression library

NVIDIA

ONNX-TensorRT

TensorRT backend for ONNX

NVIDIA

Open GPU Kernel Modules

Open-source release of CUDA GPU kernel modules for Linux

NVIDIA

Open Seq2Seq

Toolkit for efficient experimentation with various sequence-to-sequence models

NVIDIA

PhysX

Advanced, true-to-reality physics simulation engine

PyTorch

PyTorch

An open source machine learning framework that accelerates the path from research prototyping to production deployment.

RAPIDS

RAPIDS

A suite of open-source software libraries and APIs for executing data science pipelines entirely on GPUs.

NVIDIA Community

rl_games

High-performance reinforcement learning library

NVIDIA

ROS GEMs

Hardware-accelerated packages for the robot operating system (ROS)—deep neural networks (DNNs), perception pipelines, and deep learning software

NVIDIA

Sionna

An Open-Source Library for 6G Physical-Layer Research

NVIDIA

Tacotron 2

PyTorch implementation of Natural TTS Synthesis by conditioning WaveNet on Mel Spectrogram predictions

TensorFlow

TensorFlow

An Open Source Machine Learning Framework for Everyone

NVIDIA

TensorFlow-TensorRT

Integration for TensorFlow with TensorRT to get up to 6X faster inference in TensorFlow with few lines of code

NVIDIA

Thrust

Thrust is a parallel algorithms library that resembles the C++ STL.

NVIDIA

Torch-TensorRT

Integration to PyTorch with TensorRT to get 4X faster inference in PyTorch with one line of code

Pixar

Universal Scene Description (USD)

Extensible 3D scene description for composing virtual worlds

NVIDIA

vid2vid

High-resolution photorealistic video-to-video translation

NVIDIA

Warp

A Python framework that gives coders an easy way to write GPU-accelerated, kernel-based programs

AI and Deep Learning

Apache

Apache Spark

GPU-accelerated Apache Spark for data analytics, machine learning, and deep learning pipelines

NVIDIA

Apex

A PyTorch extension: Tools for easy mixed precision and distributed training in PyTorch

NVIDIA

Clara Holoscan SDK

A hybrid computing platform for medical devices, both hardware and software

NVIDIA

Collective Communication Library (NCCL)

Optimized primitives for collective multi-GPU communication

NVIDIA

Data Loading Library (DALI)

Data pre-processing in deep learning applications

NVIDIA

Deep Learning Examples

Tensor Cores-optimized code samples

NVIDIA

Deep Learning Inference Accelerator (NVDLA)

Open-source accelerator for deep learning inference

NVIDIA

Federated Learning Active Runtime Environment (FLARE)

An open-source, extensible SDK that allows researchers and data scientists to adapt existing machine learning and deep learning workflow to a privacy-preserving federated paradigm

NVIDIA

Kaolin Library

PyTorch-based library with modular optimized functionality for 3D deep learning research

NVIDIA

Merlin

Open-source framework for building high-performing recommender systems at scale

NVIDIA

MONAI

PyTorch-based framework for deep learning in healthcare imaging

NVIDIA

Morpheus

Open-source framework that enables cybersecurity developers to create optimized AI pipelines for filtering, processing, and classifying large volumes of real-time data

NVIDIA

NeMo Guardrails

A toolkit for easily developing trustworthy, safe, and secure LLM conversational systems.

NVIDIA

ONNX-TensorRT

TensorRT backend for ONNX

PyTorch

PyTorch

An open source machine learning framework that accelerates the path from research prototyping to production deployment.

RAPIDS

RAPIDS

A suite of open-source software libraries and APIs for executing data science pipelines entirely on GPUs.

NVIDIA

ROS GEMs

Hardware-accelerated packages for the robot operating system (ROS)—deep neural networks (DNNs), perception pipelines, and deep learning software

TensorFlow

TensorFlow

An Open Source Machine Learning Framework for Everyone

NVIDIA

TensorFlow-TensorRT

Integration for TensorFlow with TensorRT to get up to 6X faster inference in TensorFlow with few lines of code

NVIDIA

Torch-TensorRT

Integration to PyTorch with TensorRT to get 4X faster inference in PyTorch with one line of code

Edge AI and Robotics

NVIDIA

PhysX

Advanced, true-to-reality physics simulation engine

NVIDIA Community

rl_games

High-performance reinforcement learning library

Cloud Native Technology

NVIDIA

K8s Device Plug-in

An NVIDIA plug-in that enables GPU support in Kubernetes

NVIDIA

Cloud Native Stack

A reference architecture for managing GPU workloads with Kubernetes

NVIDIA

GPU Operator

Automates the lifecycle management of the software required to expose GPUs on Kubernetes.

NVIDIA

Network Operator

Automates the lifecycle management of the software for accelerated networking on Kubernetes.

NVIDIA

Container Toolkit

Allows users to build and run GPU accelerated containers.

Design, Visualization, and Simulation

NVIDIA

Falcor

A real-time rendering framework

NVIDIA

Material Definition Language (MDL)

Physically accurate material standard and SDK

Pixar

Universal Scene Description (USD)

Extensible 3D scene description for composing virtual worlds

NVIDIA

Warp

A Python framework that gives coders an easy way to write GPU-accelerated, kernel-based programs

High-Performance Computing (HPC)

NVIDIA

AmgX

Distributed multi-grid linear solver library on GPU

NVIDIA

CUB

A library of collective primitives and utilities

NVIDIA

CUDA Templates for Linear Algebra Subroutines (CUTLASS)

A collection of templates for high-performance general matrix multiplication (GEMM) computations within CUDA kernels

NVIDIA

CuPy

An open-source array library for GPU-accelerated computing with Python and CUDA

Flang

Flang

A Fortran compiler targeting low-level virtual machine (LLVM)

NVIDIA

libcu++

C++ standard library for an entire CPU+GPU system supported by CUDA

NVIDIA

Modulus

A neural network framework for developing Physics-ML models

NVIDIA

nvcomp

High-performance GPU data compression library

NVIDIA

Open GPU Kernel Modules

Open-source release of CUDA GPU kernel modules for Linux

NVIDIA

Thrust

Thrust is a parallel algorithms library that resembles the C++ STL.

Research

NVIDIA

FasterTransformer

Highly optimized transformer-based encoder and decoder component for GPT and BERT models

NVIDIA

Kaolin Library

PyTorch-based library with modular optimized functionality for 3D deep learning research

NVIDIA

Megatron-LM

Ongoing research for training transformer models at scale

NVIDIA

Milano

Milano is a tool for automating hyper-parameters search for your models on a backend of your choice

NVIDIA

Open Seq2Seq

Toolkit for efficient experimentation with various sequence-to-sequence models

NVIDIA

PhysX

Advanced, true-to-reality physics simulation engine

NVIDIA

Sionna

An Open-Source Library for 6G Physical-Layer Research

NVIDIA

Tacotron 2

PyTorch implementation of Natural TTS Synthesis by conditioning WaveNet on Mel Spectrogram predictions

NVIDIA

vid2vid

High-resolution photorealistic video-to-video translation

Source code for many NVIDIA Research projects is freely available for research purposes. Also, visit our SDK Glossary page for a more comprehensive list of our SDK and Library offerings.