Enterprises rely on GPU virtualization to keep their workforces productive, wherever they work. And NVIDIA virtual GPU (vGPU) performance has become essential to powering a wide range of graphics- and compute-intensive workloads from the cloud and data center. Now, designers, engineers and knowledge workers across industries can experience accelerated performance with the NVIDIA A10 and Read article >
SANTA CLARA, Calif., April 12, 2021 — GTC — NVIDIA today announced a range of eight new NVIDIA Ampere architecture GPUs for next-generation laptops, desktops and servers that make it possible…
It’s time to hail the new era of transportation. During his keynote at the GPU Technology Conference today, NVIDIA founder and CEO Jensen Huang outlined the broad ecosystem of companies developing next-generation robotaxis on NVIDIA DRIVE. These forward-looking manufacturers are set to transform the way we move with safer, more efficient vehicles for everyday mobility. Read article >
Kicking off NVIDIA’s GTC tech conference, NVIDIA CEO Jensen Huang weaves latest advancements in AI, automotive, robotics, 5G, real-time graphics, collaboration and data centers into a stunning vision of the future.
The electric vehicle revolution is about to reach the next level. Leading startups and EV brands have all announced plans to deliver intelligent vehicles to the mass market beginning in 2022. And these new, clean-energy fleets will achieve AI capabilities for greater safety and efficiency with the high-performance compute of NVIDIA DRIVE. The car industry Read article >
SANTA CLARA, Calif., April 12, 2021 (GLOBE NEWSWIRE) — NVIDIA today announced at its annual Investor Day that first quarter revenue for fiscal 2022 is tracking above its previously provided …
Data plays a crucial role in creating intelligent applications. To create an efficient AI/ ML app, you must train machine learning models with high-quality, labeled datasets. Generating and labeling such data from scratch has been a critical bottleneck for enterprises. Many companies prefer a one-stop solution to support their AI/ML workflow from data generation, data … Continued
Data plays a crucial role in creating intelligent applications. To create an efficient AI/ ML app, you must train machine learning models with high-quality, labeled datasets. Generating and labeling such data from scratch has been a critical bottleneck for enterprises. Many companies prefer a one-stop solution to support their AI/ML workflow from data generation, data labeling, model training/fine-tuning, and deployment.
To fast track the end-to-end workflow for developers, NVIDIA has been working with several partners who focus on generating large, diverse, and high-quality labeled data. Their platforms can be seamlessly integrated with NVIDIA Transfer Learning Toolkit (TLT) and NeMo for training and fine-tuning models. These efficiently trained and optimized models can then be deployed with NVIDIA DeepStream or NVIDIA Jarvis to create reliable computer vision or conversational AI applications.
Figure 1. End-to-end workflow of training a model and deploying AI applications.
In this post, we outline the key challenges in data preparation and training. We also introduce how to integrate your data to fine-tune AI/ML models easily with our partner services.
Computer vision
Large amounts of labeled data are needed to train computer vision neural network models. They could be collected from the real world or synthesized through models. High-quality labeled data enables neural network models to contextualize the information and generate accurate results.
NVIDIA integrated with the platforms from the following partners to generate synthetic and label custom data in the formats compatible with TLT for training. TLT is a zero-coding transfer learning toolkit using purpose-built, pretrained models. Trained TLT models can be deployed using DeepStream SDK and achieve 10x speedup in development time.
AI Reverie and Sky Engine
Synthetic labeled data is becoming popular especially for computer vision tasks like object detection and image segmentation. Using the platforms from AI Reverie and Sky Engine, you can generate synthetic labeled data. AI Reverie offers a suite of synthetic data for model training and validation using 3D environments which exposes the neural network models to diverse scenarios that might be hard to find in data gleaned from the real world. Sky Engine uses ray-tracing image renderer techniques to generate labeled synthetic data in virtual environments that can be used with TLT for training.
Figure 2. Computer vision workflow from unlabeled dataset to application deployment.
Appen
With Appen, you can generate and label custom data. Appen uses APIs and human intelligence to generate labeled training data. Integrated with the NVIDIA Transfer Learning Toolkit, the Appen Data Annotation platform and services allow you to eliminate time-consuming annotations and create the right training data to train with TLT for your use cases.
Hasty, Labelbox, and Sama
Few provide straightforward tools for annotation. By simple clicks and selecting the region around the objects on the image, you can quickly generate the annotations. With Hasty, Sama, and Labelbox, you can label datasets in the formats compatible with TLT. TLT is a zero-coding transfer learning toolkit using purpose-built, pretrained models. Trained TLT models can be deployed using the DeepStream SDK and achieve 10x speedup in development time.
Hasty provides such tools like DEXTR and GrabCut to create labeled data. With DEXTR, you click on the north, south, east, and west of an object and a neural network looks for the mask. For GrabCut, you select the area where the object is and add or remove regions with markings to improve the results.
With Labelbox, you can upload the data for annotation and can easily export the annotated data using Python SDK to TLT for training purposes.
Sama follows a different labeling mechanism and therefore uses a pretrained object detection model (Figure 2) to perform inference. IT uses those annotations to further enhance the labels in generating accurate results.
All these tools are tightly coupled with TLT for training to develop production-quality vision models.
Conversational AI
A typical conversational AI application may be built with automatic speech recognition (ASR), natural language understanding (NLU), dialog management (DM) and text-to-speech (TTS) services. Conversational AI models are huge and require large amounts of data for training to recognize speech in noisy environments in different languages or accents, understand every nuance of language and generate human-like speech.
NVIDIA collaborates with DefinedCrowd, LabelStudio, and DataSaur. You can use their platforms to create large and labeled datasets that can then be directly introduced into NVIDIA NeMo for training or fine-tuning. NVIDIA NeMo is an open-source toolkit for developing state-of-the-art conversational AI models.
DefinedCrowd
The popular approach for data generation/labeling is crowdsourcing. DefinedCrowd leverages a global crowd of over 500,000 contributors to collect, annotate, and validate training data by dividing the project into a series of micro-tasks through the online platform Neevo. The data is then used to train conversational AI models in a variety of languages, accents, and domains. You can now quickly download the data using the DefinedCrowd APIs and convert them into NeMo acceptable format for training.
LabelStudio
Figure 3. Speech recognition workflow from unlabeled dataset to application deployment.
Advancement of machine learning models make AI-assisted data annotation possible, which significantly reduces labeling cost. More and more vendors are integrating AI capability to augment the labeling process. NVIDIA collaborated with Heartex who maintains the open-source LabelStudio and performs AI-assisted data annotation for speech. With this tool, you can now label speech data that is compatible with NeMo (Figure 3). Heartex also integrates NeMo into their workflow so that you can choose the pretrained NeMo models for speech pre-annotation.
Datasaur
For text labeling, there are easy-to-use tools that come with predefined labels which allow you to quickly annotate the data. For more information about how to annotate your text data and run Datasaur with NeMo for training, see Datasaur x NVIDIA NeMo Integration (video).
Conclusion
NVIDIA has partnered with many industry-leading data generation and annotation providers to speed up computer vision and conversational AI application workflows. This collaboration enables you to quickly generate superior quality data that can be readily used with TLT and NeMo for training models, which can then be deployed on NVIDIA GPUs with DeepStream SDK and Jarvis.
For more information, see the following resources:
The NVIDIA HPC SDK is a comprehensive suite of compilers, libraries, and tools enabling developers to program the entire HPC platform from the GPU foundation to the CPU, and through the interconnect.
Today, NVIDIA is announcing the availability of the HPC SDK version 21.3. This software can be downloaded now free of charge.
The NVIDIA HPC SDK is a comprehensive suite of compilers, libraries, and tools enabling developers to program the entire HPC platform from the GPU foundation to the CPU, and through the interconnect. It is the only comprehensive, integrated SDK for programming accelerated computing systems.
The NVIDIA HPC SDK C++ and Fortran compilers are the first compilers to support automatic GPU acceleration of standard language constructs including C++17 parallel algorithms and Fortran intrinsics.
Learn more:
GTC 2021: S31286 A Deep Dive into the Latest HPC Software — GTC is Free, Register Now
Designing rich content and graphics for VR experiences means creating complex materials and high-resolution textures. But rendering all that content at VR resolutions and frame rates can be challenging, especially when rendering at the highest quality. You can address this challenge by using variable rate shading (VRS) to focus shader resources on certain parts of … Continued
Designing rich content and graphics for VR experiences means creating complex materials and high-resolution textures. But rendering all that content at VR resolutions and frame rates can be challenging, especially when rendering at the highest quality.
You can address this challenge by using variable rate shading (VRS) to focus shader resources on certain parts of an image—specifically, the parts that have the greatest impact for the scene, such as where the user is looking or the materials that need higher sampling rates.
However, because not all developers could integrate the full NVIDIA VRS API, NVIDIA developed Variable Rate Supersampling (VRSS), which increases the image quality in the center of the screen (fixed foveated rendering). VRSS is a zero-coding solution, so you do not need to add code to implement it. All the work is done through NVIDIA drivers, which makes it easy for users to experience VRSS in games and applications.
Video 1. Side-by-side comparison of VRSS on and off.Play the video full screen to see VRSS image quality improvement.
Figure 1. VRSS 1 fixed foveated region at 8X shading rate. Image of Boneworks courtesy of Stress Level Zero.
NVIDIA is now releasing VRSS 2, which now includes gaze-tracked, foveated rendering. The latest version further improves the perceived image quality by supersampling the region of the image where the user is looking. Applications that benefited from the original VRSS look even better with dynamic foveated rendering in VRSS 2.
New VR features
NVIDIA and Tobii collaborated to enhance VRSS with dynamic foveated rendering enabled by Tobii Spotlight, an eye-tracking technology specialized for foveation. This technology powers the NVIDIA driver with the latest eye-tracking information at minimal latency, which is used to control the supersampled region of the render frame. HPI’s upcoming Omnicept G2 HMD will be the first HMD on the market that takes advantage of this integration, as it uses both Tobii’s gaze-tracking technology and NVIDIA VRSS 2.
Figure 2. VRSS 2 with dynamic foveated region at an 8X shading rate. The blue foveated region has the highest image quality and location determined by eye gaze-tracking. Image of Boneworks courtesy of Stress Level Zero.
Video 2. VRSS 2 showing improved image quality in the green circle foveated region.Play the following video full screen to see VRSS image quality improvement. Images from Onward VR courtesy of Downpour Interactive.
Accessing VRSS in the NVIDIA control panel
VRSS has two modes in the NVIDIA Control Panel: Adaptive and Always On.
Adaptive mode
This mode takes performance limits into consideration and tries to maximize the size of the supersampling region without hindering the VR experience. The size of the foveated region grows and shrinks in proportion to the available GPU headroom.
Figure 4. Depicting the dynamic sizing of the foveated region in Adaptive mode. The supersampled region is depicted with a green mask. Supersampled region shrinks and grows based on the GPU load (scene complexity).
Always On mode
A fixed-size foveated region is always supersampled, providing maximum image quality improvements. The size of the foveated region is adequate to cover the central portion of the user’s field of view. This mode helps users perceive the maximum IQ gains possible for a given title using VRSS. However, this may also result in frame drops for applications that are performance intensive.
Figure 3. Depicting the dynamic sizing of the foveated region in Adaptive mode. Supersampled region is depicted with a green mask. Supersampled region shrinks and grows based on the GPU load (scene complexity).
Under the hood
VRSS does not require any developer integration. The entire feature is implemented and handled inside the NVIDIA display driver, provided that the application is compatible and profiled by NVIDIA.
The NVIDIA display driver handles several pieces of functionality internally:
Frame resources tracking—The driver keeps track of the resources encountered per frame and builds a heuristic model to flag a scenario where VRS could be invoked. Specifically, it notes the MSAA level to configure the VRS shading rate: the supersampling factor to be used at the center of the image. It provides render target parameters for configuring the VRS mask.
Frame render monitoring—This involves measuring the rendering load across frames, the current frame rate of the application, the target frame rate based on the HMD refresh rate, and so on. This also computes the rendering stats required for the Adaptive mode.
VRS enablement—As mentioned earlier, VRS gives you the ability to configure shading rate across the render surface. This is done with a shading rate mask and shading rate lookup table. For more information, see Turing Variable Rate Shading in VRWorks. The VRS infrastructure setup handles the configuration of the VRS mask and VRS shading rate table lookup. The VRS framework is configured based on the performance stats:
VRS mask—The size of the central region mask is configured based on the available headroom.
VRS shading rates—Configured based on the MSAA buffer sample count.
Gaze data per frame—For VRSS 2, a component supplies the latest gaze data per frame. This involves a direct data transfer between the driver and the eye-tracking vendor platform. This data is used to configure the VRS foveation mask.
Figure 4. A flow chart depicting the per-frame cycle for VRSS 2
Developerguidance
For an application to take advantage of VRSS, you must submit your game or application to NVIDIA for VRSS validation. If approved, the game or application is added to an approved list within the NVIDIA driver.
Here are the benefits for a VRSS-supported game or application:
The foveated region is now dynamic: Integration with Tobii Eye Tracking software.
Zero coding: No developer integration required to work with game or application.
Improved user experience: The user experiences VR content with added clarity.
Ease of use: User controlled on/off supersampling in the NVIDIA Control Panel.
Performance modes:Adaptive (fps priority) or Always On.
No maintenance: Technology encapsulated at the driver level.
Pavlov VR
L.A. Noire VR
Mercenary 2: Silicon Rising
Robo Recall
Eternity Warriors VR
Mercenary 2: Silicon Rising
Serious Sam VR: The Last Hope
Hot Dogs Horseshoes & Hand Grenades
Special Force VR: Infinity War
Talos Principle VR
Boneworks
Doctor Who: The Edge of time-VR
Battlewake
Lone Echo
VRChat
Job Simulator
Rec Room
PokerStars VR
Spiderman Homecoming VR
Rick & Morty Simulator: Virtual Rick-ality
Budget Cuts 2: Mission Insolvency
In Death
Skeet: VR Target Shooting
The Walking Dead: Saints & Sinners
Killing Floor: Incursion
SpiderMan far from home
Onward VR
Space Pirate Trainer
Sairento VR
Medal of Honor: Above and Beyond
The Soulkeeper VR
Raw Data
Sniper Elite VR
Table 1. 30+ game titles with NVIDIA VRSS.
Game and application compatibility
To make use of VRSS, applications must meet the following requirements:
Turing architecture
DirectX 11 VR applications
Forward rendered with MSAA
Eye-tracking software integrated with VRSS
Supersampling needs an MSAA buffer. The level of supersampling is based on the underlying number of samples used in the MSAA buffer. The foveated region is shaded 2x for 2x MSAA, 4x supersampled for 4x MSAA, and so on. The maximum shading rate applied for supersampling is 8x. The higher the MSAA level, the greater the supersampling effect will be.
Content suitability
Content that benefits from supersampling benefits from VRSS as well. Supersampling not only mitigates aliasing but also brings out details in an image. The degree of quality improvement varies across content.
Figure 5. Examples of VRSS improving the content. Image of Boneworks courtesy of Stress Level Zero.
Supersampling shines when it encounters the following types of content:
High-resolution textures
High frequency materials
Textures with alpha channels – fences, foliage, menu icons, text, and so on
Conversely, supersampling does not improve IQ for:
Flat shaded geometry
Textures and materials with low level of detail
Comeonboard!
VRSS leverages VRS technology for selective supersampling and does not require application integration. VRSS is available in NVIDIA Driver R465, and applications are required to support DX11, forward rendering, and MSAA.
If you want finer-grained control of how VRS is applied, we recommend using the explicit programming APIs of VRWorks – Variable Rate Shading (VRS). Accessing the full power of VRS, you can implement a variety of shader sampling rate optimizations, including lens-matched shading, content-adaptive shading, and gaze-tracked shading.
Magnum IO is the collection of IO technologies from NVIDIA and Mellanox that make up the IO subsystem of the modern data center and enable applications at scale. If you are trying to scale up your application to multiple GPUs, or scaling it out across multiple nodes, you are probably using some of the libraries … Continued
Magnum IO is the collection of IO technologies from NVIDIA and Mellanox that make up the IO subsystem of the modern data center and enable applications at scale. If you are trying to scale up your application to multiple GPUs, or scaling it out across multiple nodes, you are probably using some of the libraries in Magnum IO. NVIDIA is now publishing the Magnum IO Developer Environment 21.04 as an NGC container, containing a comprehensive set of tools to scale IO. This allows you to begin scaling your applications on a laptop, desktop, workstation, or in the cloud.
Magnum IO brings the ability to improve end-to-end wall clock time of IO bound applications. Imagine a workflow with three stages:
ETL (extract, transform, load)
Scaled-out compute
Post-processing and output
The first-stage ETL jobs are dominated by reading large amounts of data into GPUs and can achieve optimal performance by using Magnum IO GPUDirect Storage (GDS) for directly copying data from storage into GPU memory. This also helps reduce the CPU utilization and improves overall data center utilization.
The second stage, which comprises a distributed communication GPU-to-GPU IO intense job, can benefit from optimizing communication with NCCL message-passing or NVSHMEM shared memory models, on low latency InfiniBand networks.
The final post-processing and output stages, as well as the checkpointing and temporary storage during the workflow, can again improve performance with GDS. Magnum IO management layers also enable monitoring, troubleshooting, and detecting anomalies of all the stages of the workflow.
Figure 1. Architectural principles of Magnum IO.
Scaling applications to run efficiently is often a complex and time-consuming task. We understand that the changes to code to adopt the Magnum IO technologies can be invasive, and any changes require development, debugging, testing, and benchmarking. Magnum IO libraries also work alongside the profilers, logging, and monitoring tools needed to observe what’s happening, locate bottlenecks, and address them. You should understand the performance tradeoffs of each stage of the computation and understand the relationships and between the hardware components in the system.
Magnum IO libraries provide APIs that manage the underlying hardware, allowing you to focus on the algorithmic aspects of your applications. The APIs are designed to be high-level so that they are easy to integrate with, but also to expose finer controls for when you start fine-tuning performance, after the behaviors and tradeoffs of the application running at scale are understood.
The high bandwidth and low latency offered by NVLink operating at 300GB/s, and InfiniBand in NVIDIA DGX A100 systems also opens new possibilities for algorithms. The NVLink bandwidth between GPUs now makes remote memory almost local. The total number of PCIe lanes from remote storage may exceed those from local storage. Magnum IO libraries on NVIDIA hardware allow algorithms to take full advantage of the GPU memory across all nodes, rather than sacrificing efficiency to avoid what was bottlenecking IO with high latencies in the past.
Figure 2. Magnum IO technologies.
Magnum IO: GPU-to-GPU communications
Core to Magnum IO are libraries that allow GPUs to talk directly to each other over the fasted links available.
NCCL
The NVIDIA Collective Communications Library (NCCL, pronounced “nickel”) is a library providing inter-GPU communication primitives that are topology-aware and can be easily integrated into applications.
NCCL is smart about IO on systems with complex topology: systems with multiple CPUs, GPUs, PCI busses, and network interfaces. It can selectively use NVLink, Ethernet, and InfiniBand, using multiple links when possible. Consider using NCCL APIs whenever you plan your application or library to run on a mix of multi-GPU multi-node systems in a data center, cloud, or hybrid system. At runtime, NCCL determines the topology and optimizes layout and communication methods.
NVSHMEM
NVSHMEM creates a global address space for data that spans the memory of multiple GPUs and can be accessed with fine-grained GPU-initiated operations, CPU-initiated operations, and operations on CUDA streams.
In many HPC workflows, models and simulations are run that far exceed the size of a single GPU or node. NVSHMEM allows for a simpler asynchronous communication model in a shared address space that spans GPUs within or across nodes, with lower overheads, possibly resulting in stronger scaling compared to a traditional Message Passing Interface (MPI).
UCX
Unified Communication X (UCX) uses high-speed networks, including InfiniBand, for inter-node communication and shared memory mechanisms for efficient intra-node communication. If you need a standard CPU-driven MPI, PGAS OpenSHMEM libraries, and RPC, GPU-aware communication is layered on top of UCX.
UCX is appropriate when driving IO from the CPU, or when system memory is being shared. UCX enables offloading the IO operations to both host adapter (HCA) and switch, which reduces CPU load. UCX simplifies the portability of many peer-to-peer operations in MPI systems.
Magnum IO: Storage-to-GPU communications
Magnum IO also addresses the need to move data between GPUs and storage systems, both local and remote, with as little overhead as possible along the way.
GDS
NVIDIA GPUDirect Storage (GDS) enables a direct data path for Remote Direct Memory Access (RDMA) transfers between GPU memory and storage, which avoids a bounce buffer and management by the CPU. This direct path increases system bandwidth and decreases the latency and utilization load on the CPU.
a) Without GDS
b) With GDS
Figure 3. GDS avoids the bounce buffer on CPU and achieves higher IO bandwidth, lower latency, and access to O(PB) capacity, all enabled with a simple CUDA programming model.
GDS and the cuFile APIs should be used whenever data needs to move directly between storage and the GPU. With storage systems that support GDS, significant increases in performance on clients are observed when IO is a bottleneck. In cases where the storage system does not support GDS, IO transparently falls back to normal file reads and writes.
Moving the IO decode/encode from the CPU to GPU creates new opportunities for direct data transfers between storage and GPU memory which can benefit from GDS performance. An increasing number of data formats are supported in CUDA.
Magnum IO: Profiling and optimization
NVIDIA Nsight Systems lets you see what’s happening in the system and NVIDIA Cumulus NetQ allows you to analyze what’s happening on the NICs and switches. This is critical to finding some causes of bottlenecks in multi-node applications.
Nsight Systems
Nsight Systems is a low-overhead performance analysis tool designed to provide insights that you need to optimize your software. It provides everything that you would expect from a profiler for a GPU. Nsight Systems has a tight integration with many core CUDA libraries, giving you detailed information on what is happening.
Nsight Systems allows you to see exactly what’s happening on the system, what code is taking a long time, and when algorithms are waiting on GPU/CPU compute, or device IO. Nsight Systems is relevant to Magnum IO and included in the Magnum IO container for convenience, but its scope spans well outside of Magnum IO to monitoring compute that’s unrelated to IO.
Figure 4. Diagram of a NCCL application trace with Nsight Systems.
NetQ
NetQ is a highly scalable, modern, network operations tool set that provides visibility, troubleshooting and lifecycle management of your open networks in real time. It enables network profiling functionality that can be used along with Nsight Systems or application logs to observe the network’s behavior while the application is running.
NetQ is part of Magnum IO itself, given its integral involvement in managing IO in addition to profiling it.
Getting started with the Magnum IO Developer Environment
We are launching the Magnum IO Developer Environment as a container hosted on NVIDIA NGC for GTC 21. A bare-metal installer for Ubuntu and RHEL may be coming soon. The container provides a sealed environment with the latest versions of the libraries compatible with each other. It makes it easy for you to begin optimizing your application’s IO. Installing and working with the container does not interfere with any existing system setup, which may have different versions of the components.
The Magnum IO components included in the 21.04 container are as follows:
Ubuntu 20.04
CUDA
Nsight Systems CLI
GDS
GPUDirect RDMA
GPUDirect P2P
NCCL
UCX
NVSHMEM
The first step is to profile the application and find the bottlenecks, then evaluate which of the Magnum IO tools, libraries, or algorithm changes are appropriate for removing those bottlenecks and optimizing the application.
Download the developer environment today: Magnum IO SDK.
Data plays a crucial role in creating intelligent applications. To create an efficient AI/ ML app, you must train machine learning models with high-quality, labeled datasets. Generating and labeling such data from scratch has been a critical bottleneck for enterprises. Many companies prefer a one-stop solution to support their AI/ML workflow from data generation, data … 
The NVIDIA HPC SDK is a comprehensive suite of compilers, libraries, and tools enabling developers to program the entire HPC platform from the GPU foundation to the CPU, and through the interconnect. 
Designing rich content and graphics for VR experiences means creating complex materials and high-resolution textures. But rendering all that content at VR resolutions and frame rates can be challenging, especially when rendering at the highest quality. You can address this challenge by using variable rate shading (VRS) to focus shader resources on certain parts of … 
Magnum IO is the collection of IO technologies from NVIDIA and Mellanox that make up the IO subsystem of the modern data center and enable applications at scale. If you are trying to scale up your application to multiple GPUs, or scaling it out across multiple nodes, you are probably using some of the libraries … 
