The NVIDIA Deep Learning Examples repository, introduced in 2019, offers state-of-the-art Deep Learning (DL) scripts across various domains, including Computer Vision, Natural Language Processing (NLP), and Recommender Systems. These examples are optimized for NVIDIA GPUs, ensuring high performance and reproducibility in training and deployment. Developers can leverage these resources to build and refine their own DL models efficiently.
Launched in 2019, NVIDIA NeMo is a scalable Generative AI framework tailored for researchers and developers working on Large Language Models (LLMs), multimodal applications, and speech AI systems. It provides tools for building, training, and fine-tuning models in Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and other NLP tasks, facilitating the development of sophisticated AI applications.
Introduced in 2021, NVIDIA Modulus is an open-source Deep Learning framework designed for developing physics-informed AI models. It combines data-driven methods with physics-based principles to deliver real-time predictions in scientific and engineering applications, enabling accurate simulations and analyses.
NVIDIA Merlin, released in 2020, is an open-source library aimed at building high-performance, end-to-end Recommender Systems. It encompasses components for data preprocessing, training, and deployment, optimized for GPU acceleration to handle large-scale recommendation tasks efficiently.
Megatron-LM, developed by NVIDIA in 2019, is a framework for training large-scale Transformer models. It employs model parallelism techniques to facilitate the training of models with billions of parameters across multiple GPUs, advancing the capabilities of LLMs.
The NVIDIA Data Loading Library (DALI), introduced in 2018, is a GPU-accelerated library for data preprocessing in Deep Learning applications. It provides highly optimized building blocks and an execution engine to accelerate data pipelines, enhancing training efficiency.
NVIDIA TensorRT, launched in 2017, is a high-performance deep learning inference library. It optimizes trained models for deployment, delivering low latency and high throughput for AI applications across various platforms.
NVIDIA Clara Train, released in 2019, is a Deep Learning framework for medical imaging. It offers tools for training and deploying AI models in healthcare, facilitating the development of applications like image segmentation and classification.
Introduced in 2019, NVIDIA Kaolin is a PyTorch library aimed at accelerating 3D deep learning research. It provides efficient implementations of 3D data structures and algorithms, enabling researchers to develop and evaluate 3D Deep Learning models effectively.
NVIDIA Apex, launched in 2018, is a PyTorch extension that facilitates mixed precision training and distributed training. It helps developers optimize memory usage and computational efficiency, enabling the training of larger models and faster experimentation.
https://github.com/NVIDIA/apex
NVIDIA RAPIDS, introduced in 2018, is an open-source suite of data processing and Machine Learning (ML) libraries that leverages GPU acceleration. It enables end-to-end data science workflows, including data preparation, model training, and inference, all within a Python ecosystem.
NVIDIA PyTorch Lightning Plugins, launched in 2020, extend the capabilities of PyTorch Lightning for Deep Learning (DL) model training. These plugins optimize multi-GPU and distributed training, enabling developers to scale their models efficiently on NVIDIA GPUs.
https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/Training/PyTorchLightning
NVIDIA Triton Inference Server, introduced in 2019, provides a scalable and efficient platform for deploying Deep Learning (DL) models in production. It supports multiple frameworks, including TensorFlow, PyTorch, and ONNX Runtime, delivering high-performance inference on NVIDIA GPUs.
NVIDIA Maxine, launched in 2020, is a suite of AI technologies for real-time video and voice communication. It provides features like noise reduction, face alignment, and virtual backgrounds, enhancing virtual meeting experiences through GPU-accelerated AI processing.
NVIDIA StyleGAN2, released in 2020, builds on the original StyleGAN architecture for generating high-quality, photorealistic images. It is widely used in generative art, gaming, and AI research, showcasing state-of-the-art capabilities in Generative Adversarial Networks (GANs).
NVIDIA FLARE, introduced in 2021, is a federated learning framework for secure and collaborative AI training. It allows multiple organizations to train models on decentralized data without compromising privacy or data security, leveraging NVIDIA GPUs for efficiency.
NVIDIA DeepStream, launched in 2018, is a development framework for video analytics and AI applications at the edge. It integrates Computer Vision models with video pipelines, enabling real-time analytics in industries like retail, transportation, and security.
NVIDIA DIGITS, introduced in 2015, is a graphical interface for training and visualizing Deep Learning models. It simplifies data ingestion, network design, and model evaluation, making it ideal for beginners and researchers in AI.
NVIDIA Clara Parabricks, released in 2019, provides tools for genomic data analysis using AI and GPU acceleration. It significantly reduces processing time for genomics pipelines, enabling faster insights in medical and research applications.
NVIDIA Jetson Inference, launched in 2018, is an open-source library for deploying AI models on NVIDIA Jetson devices. It includes pre-trained models for object detection, image classification, and segmentation, optimized for edge AI applications.
https://github.com/dusty-nv/jetson-inference
NVIDIA DLSS, introduced in 2018, leverages Deep Learning (DL) to improve gaming performance and visual quality. It uses AI to upscale lower-resolution frames to higher resolutions, providing smoother gameplay and better graphics fidelity on NVIDIA GPUs.
NVIDIA CUDA-X AI, launched in 2019, is a collection of GPU-accelerated libraries for Deep Learning (DL), Machine Learning (ML), and Data Science. It integrates seamlessly with CUDA to deliver high-performance computing for AI applications.
NVIDIA JetBot, introduced in 2019, is an open-source robotics platform based on NVIDIA Jetson devices. It enables developers and students to explore AI and robotics by implementing Computer Vision and autonomous navigation projects.
NVIDIA OpenSeq2Seq, released in 2018, is an open-source toolkit for training and deploying sequence-to-sequence models. It supports tasks like Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and Natural Language Processing (NLP).
NVIDIA Isaac SDK, launched in 2019, is a comprehensive platform for developing AI-powered robotics applications. It provides simulation tools, pre-trained models, and real-world deployment capabilities for autonomous robots.
NVIDIA Kaolin Wisp, introduced in 2021, is an extension of the NVIDIA Kaolin library for volumetric rendering and implicit neural representations. It focuses on accelerating research in 3D Deep Learning (DL).
NVIDIA Torch-TensorRT, launched in 2020, is a PyTorch extension for optimizing Deep Learning (DL) inference using TensorRT. It accelerates model deployment on NVIDIA GPUs, delivering high throughput and low latency.
NVIDIA Clara AGX, released in 2020, is a platform for AI medical devices. It integrates GPU-accelerated computing and Deep Learning tools to enable real-time diagnostics and treatment planning in healthcare.
NVIDIA Morpheus, introduced in 2021, is a cybersecurity AI framework for analyzing and securing large-scale network data. It uses Deep Learning (DL) and NVIDIA GPUs to detect anomalies and respond to threats in real time.
NVIDIA DOCA (Data-Center Optimized Computing Acceleration), launched in 2021, is a software framework for building AI and Data Science applications on NVIDIA BlueField DPUs. It accelerates data-center operations and enhances security.
https://github.com/NVIDIA/doca-sdk
NVIDIA Omniverse, introduced in 2021, is a real-time 3D collaboration and simulation platform. It leverages AI and GPU acceleration for creating, rendering, and simulating complex virtual environments, supporting industries like architecture, media, and gaming.
NVIDIA Clara Imaging, launched in 2018, provides AI tools and pre-trained models for medical imaging. It supports image segmentation, classification, and reconstruction tasks, enabling advancements in diagnostics and research.
NVIDIA Riva, released in 2020, is a framework for building and deploying AI-powered conversational applications. It provides tools for Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU) optimized for real-time use cases.
NVIDIA Transfer Learning Toolkit, introduced in 2019, simplifies the customization of pre-trained AI models for specific applications. It supports transfer learning for Computer Vision and Natural Language Processing (NLP) tasks.
NVIDIA cuML, part of the RAPIDS suite, is a GPU-accelerated library for Machine Learning (ML). Launched in 2018, it provides fast implementations of common ML algorithms like clustering, regression, and dimensionality reduction.
NVIDIA cuGraph, also part of RAPIDS, focuses on Graph Analytics using GPU acceleration. Introduced in 2019, it supports algorithms like PageRank, community detection, and shortest path computations for large-scale graph data.
NVIDIA Maxine AR SDK, launched in 2020, provides tools for creating augmented reality applications. It uses AI for real-time face tracking, animation, and video enhancement, supporting immersive digital experiences.
NVIDIA JetPack, introduced in 2014, is a software development kit for NVIDIA Jetson devices. It includes libraries and tools for deploying AI applications on edge devices, supporting industries like robotics, healthcare, and IoT.
NVIDIA Clara Holoscan, launched in 2021, is a platform for AI-powered medical devices and real-time imaging. It supports high-throughput image processing and diagnostics, enabling next-generation healthcare solutions.
NVIDIA WaveGlow, introduced in 2018, is a Deep Learning model for Text-to-Speech (TTS). It generates high-quality audio waveforms directly from text, enabling natural and expressive speech synthesis.