Categories
Game

HBO releases its first ‘The Last of Us’ trailer

September 26th is The Last Of Us Day and HBO is marking the occasion with the first trailer for its highly anticipated adaptation of Naughty Dog’s game. The clip, which runs for just over a minute and a half, includes some of the most memorable moments from the 2013 title (which just got a complete remake for PS5). It features the leaning building from an early level and a bombastic opening outbreak sequence, as well as the unmistakable sound of a Clicker. It seems that the show will draw from the Left Behind expansion as well.

For the uninitiated, The Last of Us tells the story of Joel (Pedro Pascal) and Ellie (Bella Ramsay). The pair travel across a near-future version of the US that has been left devastated by a fungal infection, which turns victims into aggressive, zombie-like creatures. The trailer does a solid job of capturing the terrifying atmosphere of the game.

It’s clear HBO has high hopes for the series. Last month, it closed out a big sizzle reel that highlighted upcoming projects with the first footage from the show. While there’s no specific release date as yet, The Last of Us will premiere on HBO and HBO Max in early 2023.

All products recommended by Engadget are selected by our editorial team, independent of our parent company. Some of our stories include affiliate links. If you buy something through one of these links, we may earn an affiliate commission. All prices are correct at the time of publishing.

Repost: Original Source and Author Link

Categories
Game

CD Projekt Red releases an official modding tool for ‘Cyberpunk 2077’

Cyberpunk 2077 now has an official modding tool. CD Projekt RED has launched REDmod, which provides players integrated support to easily install and load mods onto the PC version of the action RPG. As the developer’s official announcement notes, it will also allow players to modify and personalize their game by using the custom sounds, animations and scripts that come with the tool. CD Projekt Red promises to update the tool alongside future patches to ensure that it remains compatible with the game. It is a free DLC, though, and players don’t have to install it at all if they don’t want to.

As popular mod website Nexus Mods clarifies, while new mods are required to use a specific format to be compatible with REDmod, old mods will continue to work just fine. Older mods that aren’t compatible with the tool simply won’t show up in the new REDmod menu. That’s also were players can toggle mods that are compatible with the tool on or off. 

The free DLC is now available for download from the official Cyberpunk 2077 website, but players can get also get it from GOG, Steam or Epic.

All products recommended by Engadget are selected by our editorial team, independent of our parent company. Some of our stories include affiliate links. If you buy something through one of these links, we may earn an affiliate commission. All prices are correct at the time of publishing.

Repost: Original Source and Author Link

Categories
AI

MLCommons releases new benchmarks to boost ML performance

Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.


Understanding the performance characteristics of different hardware and software for machine learning (ML) is critical for organizations that want to optimize their deployments.

One of the ways to understand the capabilities of hardware and software for ML is by using benchmarks from MLCommons — a multi-stakeholder organization that builds out different performance benchmarks to help advance the state of ML technology. 

The MLCommons MLPerf testing regimen has a series of different areas where benchmarks are conducted throughout the year. In early July, MLCommons released benchmarks on ML training data and today is releasing its latest set of MLPerf benchmarks for ML inference. With training, a model learns from data, while inference is about how a model “infers” or gives a result from new data, such as a computer vision model that uses inference for image recognition.

The benchmarks come from the MLPerf Inference v2.1 update, which introduces new models, including SSD-ResNeXt50 for computer vision, and a new testing division for inference over the network to help expand the testing suite to better replicate real-world scenarios.

Event

MetaBeat 2022

MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.

Register Here

“MLCommons is a global community and our interest really is to enable ML for everyone,” Vijay Janapa Reddi, vice president of MLCommons said during a press briefing. “What this means is actually bringing together all the hardware and software players in the ecosystem around machine learning so we can try and speak the same language.”

He added that speaking the same language is all about having standardized ways of claiming and reporting ML performance metrics.

How MLPerf measures ML inference benchmarks

Reddi emphasized that benchmarking is a challenging activity in ML inference, as there are any number of different variables that are constantly changing. He noted that MLCommons’ goal is to measure performance in a standardized way to help track progress.

Inference spans many areas that are considered in the MLPerf 2.1 suite, including recommendations, speech recognition, image classification and object detection capabilities. Reddi explained that MLCommons pulls in public data, then has a trained ML network model for which the code is available. The group then determined a certain target quality score that submitters of different hardware systems platforms need to meet.

“Ultimately, our goal here is to make sure that things get improved, so if we can measure them, we can improve them,” he said.

Results? MLPerf Inference has thousands

The MLPerf Inference 2.1 suite benchmark is not a listing for the faint of heart, or those that are afraid of numbers — lots and lots of numbers.

In total the new benchmark generated over 5,300 results, provided by a laundry list of submitters including Alibaba, Asustek, Azure, Biren, Dell, Fujitsu, Gigabyte, H3C, HPE, Inspur, Intel, Krai, Lenovo, Moffett, Nettrix, NeuralMagic, Nvidia, OctoML, Qualcomm, Sapeon and Supermicro.

“It’s very exciting to see that we’ve got over 5,300 performance results, in addition to over 2,400 power measurement results,” Reddi said. “So there’s a wealth of data to look at.”

The volume of data is overwhelming and includes systems that are just coming to market. For example, among Nvidia’s many submissions are several for the company’s next generation H100 accelerator that was first announced back in March.

“The H100 is delivering phenomenal speedups versus previous generations and versus other competitors,” Dave Salvator, director of product marketing at Nvidia, commented during a press briefing that Nvidia hosted.

While Salvator is confident in Nvidia’s performance, he noted that from his perspective it’s also good to see new competitors show up in the latest MLPerf Inference 2.1 benchmarks. Among those new competitors is Chinese artificial intelligence (AI) accelerator vendor Biren Technology. Salvator noted that Biren brought in a new accelerator that he said made a “decent” first showing in the MLPerf Inference benchmarks.

“With that said, you can see the H100 outperform them (Biren) handily and the H100 will be in market here very soon before the end of this year,” Salvator said.

Forget about AI hype, enterprises should focus on what matters to them

The MLPerf Inference numbers, while verbose and potentially overwhelming, also have a real meaning that can help to cut through AI hype, according to Jordan Plawner, senior director of Intel AI products.

“I think we probably can all agree there’s been a lot of hype in AI,” Plawner commented during the MLCommons press briefing. “I think my experience is that customers are very wary of PowerPoint in claims or claims based on one model.”

Plawner noted that some models are great for certain use cases, but not all use cases. He said that MLPerf helps him and Intel communicate to customers in a credible way with a common framework that looks at multiple models. While attempting to translate real-world problems into benchmarks is an imperfect exercise, MLPerf has a lot of value.

“This is the industry’s best effort to say here [is] an objective set of measures to at least say — is company XYZ credible,” Plawner said.

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.

Repost: Original Source and Author Link

Categories
Computing

Lenovo releases two new ThinkPad workstations

Lenovo has just unveiled two new ThinkPad workstations during SIGGRAPH 2022. The range includes the ThinkPad P15v and the ThinkPad P14s, both of which are laptops.

These new mobile workstations will come equipped with some of the latest hardware from AMD and Nvidia, including AMD Ryzen Pro CPUs and Nvidia’s workstation RTX graphics.

Lenovo

Lenovo’s ThinkPads scarcely need an introduction. The company continues expanding its workstation presence by adding to it every so often, and now, two new models are joining the lineup. Lenovo markets the new releases toward creators, touting the fact that they’re more value-oriented than some of the high-end solutions it also offers.

Made to serve as a way to bridge the gap between the entry-level and premium segments, the two new ThinkPads vary in terms of both specs and size. The P14s is a smaller laptop, and according to Lenovo, it’s also the company’s lightest mobile workstation. With a 14-inch chassis and a weight of less than 3 pounds, it’s certainly going to be a slim, portable device.

The ThinkPad P14s can be configured with an AMD Ryzen Pro U-series CPU, including Ryzen 5 and Ryzen 7 options, with up to eight cores and a maximum clock speed of 4.7GHz. It doesn’t come with a discrete GPU, however, so it may do a better job in less graphics-intensive workloads — but with an integrated Radeon card, it will still suffice for many use cases.

Moving on to the ThinkPad P15v, this is the bigger and better of the two siblings, equipped with discrete graphics and a larger display. It weighs in at just under 5 pounds. It comes with a 15.6-inch display and a maximum resolution of UHD, meaning 3840 x 2160, although there’s also an FHD option at 1920 x 1080. Both screens feature IPS panels for increased brightness and better color reproduction.

Engineer, wearing a hard hat, works on the Lenovo ThinkPad P14s as another engineer works in the background.
Lenovo

Processor options are similar, but here we have AMD Ryzen Pro H-series CPUs, including Ryzen 5 and Ryzen 7 variants with the same core configuration and clocks as the P14s. Customers will be able to pick out an up to Nvidia RTX A2000 graphics card, making this one a strong option for content creators and various creatives.

Storage and memory options include up to 32GB of LPDDR5-6400 RAM for the P14s and a maximum of 64GB DDR5-4800 for the P15v, as well as up to 2TB storage for the smaller workstation and up to 4TB for the bigger version. You can pick either Windows 11 Pro or Home, Windows 10 Pro, or Linux.

It’s unclear when exactly the new workstations will be up for sale or how much they will cost. Each can be configured to match your specific needs, so the price will surely vary based on the specs that are picked.

Editors’ Choice




Repost: Original Source and Author Link

Categories
AI

Dataiku releases new version of unified AI platform for machine learning

Hear from CIOs, CTOs, and other C-level and senior execs on data and AI strategies at the Future of Work Summit this January 12, 2022. Learn more


Dataiku recently released version 10 of its unified AI platform. VentureBeat talked to Dan Darnell, head of product marketing at Dataiku and former VP of product marketing at H2O.ai, to discuss how the new release provides greater governance and oversight of the enterprise’s machine learning efforts, enhances ML ops, and enables enterprises to scale their ML and AI efforts.

Governance and oversight

For Darnell, the name of the game is governance. “Until recently,” he told VentureBeat, “data science tooling at many enterprises has been the wild west, with different groups adopting their favorite tools.” However, he sees a noticeable change in tooling becoming consolidated “as enterprises are realizing they lack visibility into these siloed environments, which poses a huge operational and compliance risk. They are searching for a single ML repository to provide better governance and oversight.” Dataiku is not alone in spotting this trend, with competing products like AWS MLOps tackling the same space.

Having a single point of governance is helpful for enterprise users. Darnell likens it to a single “watchtower, from which to view all of an organization’s data projects.” For Dataiku, this enables project workflows that provide blueprints for projects, approval workflows that require managerial sign-off before deploying new models, risk and value assessment to score their AI projects, and a centralized model registry to version models and track model performance.

For its new release, governance is centered around the “project,” which also contains the data sources, code, notebooks, models, approval rules, and markdown wikis associated with that effort. Just as GitHub went beyond mere code hosting to hosting the context around coding that facilitates collaboration, such as pull requests, CI/CD, markdown wikis, and project workflow, Dataiku‘s eponymous “projects” aspire to do the same for data projects. “Whether you write your model inside Dataiku or elsewhere, we want you to put that model into our product,” said Darnell.

ML ops

Governance and oversight also extend into the emerging field of ML ops, a rapidly growing discipline that applies several DevOps best practices for machine learning models. In its press release, Dataiku defines ML ops as helping “IT operators and data scientists evaluate, monitor and compare machine learning models, whether under development or in production.” In this area, Dataiku competes against products like Sagmaker’s Model Monitor, GCP’s Vertex AI Model Monitoring, or Azure’s MLOps.

Automatic drift analysis is an important newly released feature. Over time, data can fluctuate due to subtle underlying changes outside the modeler’s control. For example, as the pandemic progressed and consumers began to see delays in gym re-openings, sales of home exercise equipment began creeping up. This data drift can lead to poor performance for models that were trained on out-of-date data.

What-If scenarios are one of the more interesting features of the new AI platform. Machine learning models usually live in code, accessible only to trained data scientists, data engineers, and the computer systems that process them. But nontechnical business stakeholders want to see how the model works for themselves. These domain experts often have significant knowledge, and they often want to get comfortable with a model before approving it. Dataiku what-if “simulations” wrap a model so that non-technical stakeholders can interrogate the model by setting different inputs in an interactive GUI, without diving into the code. “Empowering non-technical users as part of the data science workflow is a critical component of MLOps,” Darnell said.

Scaling ML and AI

“We think that ML and AI will be everywhere in the organization, and we have to unlock the bottleneck of the data scientist being the only person who can do ML work,” Darnell said.

One way Dataiku is tackling it is to reduce the duplicative work of data scientists and analysts. Duplicative work is the bane of any large enterprise where code silos are rampant. Data scientists redo the work because they simply don’t know if it was done elsewhere. A catalog of code snippets can provide data scientists and analysts greater visibility on prior work so that they can stand on the shoulders of colleagues rather than reinvent the wheel. Whether or not the catalog can work will hinge on search performance — a notoriously tricky problem — as well as whether search can easily identify the relevant prior work, therefore freeing up data scientists to accomplish more valuable tasks.

In addition to trying to make data scientists more effective, Dataiku’s AI platform also provides no-code GUIs for data prep and AutoML capabilities to perform ETL, train models, and assess their quality. This feature is geared at technically-proficient users who cannot code and empowers them to do many of the data science tasks. Through a no-code GUI, users can control which ML models are available to the AutoML algorithm and perform basic feature manipulations on the input data. After training, the page provides visuals to aid in model interpretability, not just regression coefficients, hyperparameter selection, and performance metrics, but more sophisticated diagnostics like subpopulation analysis. The latter is very helpful for AI bias, where model performance may be very strong overall but weak for a vulnerable subpopulation, leading to bias. No-code solutions are hot, with AWS also releasing Sagemaker Canvas, a competing product.

More on Dataiku

Dataiku’s initial product, the “Data Science Studio,” focused on providing tooling for the individual data scientist to become more productive. With Dataiku 10, its focus is shifted to the enterprise, with features that target the CTO as well as the rank and file data scientist. This shift is not uncommon among data science vendors chasing stickier seven-figure enterprise deals with higher investor multiples. This direction mirrors similar moves by well-established competitors in the cloud enterprise data science space, including Databricks, Oracle’s Autonomous DataWarehouse, GCP Vertex, Microsoft’s Azure ML, and AWS Sagemaker, which VentureBeat has written about previously.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
AI

Google releases TF-GNN for creating graph neural networks in TensorFlow Google has released

Hear from CIOs, CTOs, and other C-level and senior execs on data and AI strategies at the Future of Work Summit this January 12, 2022. Learn more


Let the OSS Enterprise newsletter guide your open source journey! Sign up here.

Google today released TensorFlow Graph Neural Networks (TF-GNN) in alpha, a library designed to make it easier to work with graph structured data using TensorFlow, its machine learning framework. Used in production at Google for spam and anomaly detection, traffic estimation, and YouTube content labeling, Google says that TF-GNN is designed to “encourage collaborations with researchers in industry.”

Graphs are a set of objects, places, or people and the connections between them. A graph represents the relations (edges) between a collection of entities (nodes or vertices), all of which can store data. Directionality can be ascribed to the edges to describe information, traffic flow, and more.

More often than not, the data in machine learning problems is structured or relational and thus can be described with a graph. Fundamental research on GNNs is decades old, but recent advances have led to great achievements in many domains, like modeling the transition of glass from a liquid to a solid and predicting pedestrian, cyclist, and driver behavior on the road.

TF-GNN

Above: Graphs can model the relationships between many different types of data, including web pages (left), social connections (center), or molecules (right).

Image Credit: Google

Indeed, GNNs can be used to answer questions about multiple characteristics of graphs. By working at the graph level, they can try to predict aspects of the entire graph, for example identifying the presence of certain “shapes” like circles in a graph that might represent close social relationships. GNNs can also be used on node-level tasks to classify the nodes of a graph or at the edge level to discover connections between entities.

TF-GNN

TF-GNN provides building blocks for implementing GNN models in TensorFlow. Beyond the modeling APIs, the library also delivers tooling around the task of working with graph data, including a data-handling pipeline and example models.

Also included with TF-GNN is an API to create GNN models that can be composed with other types of AI models. In addition to this, TF-GNN ships with a schema to declare the topology of a graph (and tools to validate it), helping to describe the shape of training data.

“Graphs are all around us, in the real world and in our engineered systems … In particular, given the myriad types of data at Google, our library was designed with heterogeneous graphs in mind,” Google’s Sibon Li, Jan Pfeifer, Bryan Perozzi, and Douglas Yarrington wrote in the blog post introducing TF-GNN.

TF-GNN adds to Google’s growing collection of TensorFlow libraries, which spans TensorFlow Privacy, TensorFlow Federated, and TensorFlow.Text. More recently, the company open-sourced TensorFlow Similarity, which trains models that search for related items — for example, finding similar-looking clothes and identifying currently playing songs.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
Game

Nintendo releases big ‘Animal Crossing: New Horizons’ update earlier than expected

Nintendo has released its last free major content update for Animal Crossing: New Horizons over a day earlier than planned. During its Direct presentation in October, the gaming giant announced that it’s rolling out New Horizons version 2.0 on November 5th. As TechCrunch and IGN have confirmed, though, the update is now live and can be downloaded to your Switch. Version 2.0 adds quite a number of new features to the game, including characters from old Animal Crossing titles. 

One of those characters is Brewster, the quiet pigeon proprietor who’ll open up the Roost café at the museum after you do a certain favor for museum director Blathers. Kapp’n, the singing sailor kappa, is also back and will take you to remote islands on his boat. You can only purchase boat rides once a day with Nook Miles, though, so you can’t endlessly sail around all day. If you want to shop from new stores owned by familiar characters, you can head over to Harv’s Island, which now has an open market. Reese & Cyrus’ shop, for instance, will offer new types of furniture customization, while Katrina will read your fortune.

The update adds gyroid hunting and cooking activities, as well. For the latter, which will be part of DIY recipes, you can combine anything you harvest and other ingredients to create new dishes. Finally, New Horizons 2.0 introduces several quality-of-life improvements, including the ability to establish ordinances. You can make the residents get up at the time of the day you’re active in the game, for example, or reduce weeds’ growing rate. The update also allows you to keep more items by giving you a bigger home storage and storage sheds you can place around your island. 

In addition to the free update, Nintendo announced last month that it’s releasing a Happy Home Paradise paid DLC that’ll let you design vacation homes for characters on November 5th. That one isn’t available yet, but it’ll set you back $25 when it comes out tomorrow. 

All products recommended by Engadget are selected by our editorial team, independent of our parent company. Some of our stories include affiliate links. If you buy something through one of these links, we may earn an affiliate commission.

Repost: Original Source and Author Link

Categories
AI

Nvidia releases robot toolbox to deepen support of AI-powered robotics in ROS

Nvidia announced today that Isaac, its developer toolbox for supporting AI-powered robotics, will deepen support of the Robot Operating System (ROS).

The announcement is being made this morning at ROS World 2021, a conference for developers, engineers, and hobbyists who work on ROS, a popular open-source framework that helps developers build and reuse code used for robotics applications.

Nvidia, which is trying to assert its lead as a supplier of processors for AI applications, announced a host of “performance perception” technologies that would be part of what it will now call Isaac ROS. This includes computer vision and AI/ML functionality in ROS-based applications to support things like autonomous robots.

The move comes as Amazon’s robotic platform, RoboMaker, has also moved quickly to support ROS.

The ROS World 2021 is the ninth annual developers’ conference — modeled after PyCon and BoostCon — for developers of all levels to learn from and network with the ROS community.

Nvidia said its offerings are intended to accelerate and improve the standards of product development and product performance.

Isaac ROS GEM solution for optimized real-time Stereo Visual Odometry Solution

The purpose of the newly launched Isaac ROS GEM for Stereo Visual Odometry is to help autonomous vehicles keep track of where a camera is relative to its initial position. If seen from a broader perspective, it assists these autonomous machines to track where they are concerning the larger environment.

With this solution, ROS developers get a real-time (>60fps@720p) stereo camera visual odometry solution that runs immensely fast and can run HD resolution in real-time on a Jetson Xavier AGX.

ROS developers can now access all Nvidia NGC DNN inference models

With DNN Inference GEM, ROS developers can now leverage any of Nvidia’s inference models available on NGC, or can offer their own DNN. TensorRT or Triton, Nvidia’s inference servers, will deploy these optimized packages. The GEM is also compatible with U-Net and DOPE. The U-Net helps generate semantic segmentation masks from images, while DOPE helps in estimating three-dimensional poses for all detected objects. If you are keen to integrate performant AI inference in a ROS application, the DNN inference GMM is one of the fastest alternatives you can get.

Isaac SIM GA release for AI-powered robotics

Scheduled to be launched in November 2021, this GA release of Isaac SIM will come with improvements in the UI and performance, making simulation-building much faster. The ROS bridge will improve, and so will the developer experience with an increased number of ROS samples. The new release will reduce memory usage and startup times and better the process of Occupancy Map Generation. The new environment variants include large warehouses, offices, and hospitals, and the new Python building blocks can interface with robots, objects, and environments.

Synthetic data generation workflow

Addressing the safety and quality concerns of autonomous robots is crucial as it deals with a large and diverse data volume to shape up its AI models perfectly. It is these AI models that run the perception stack. The new synthetic data workflow that comes with the Isaac Sim helps build production quality datasets, addressing the safety and quality concerns of autonomous robots.

With this data generation workflow, the control of the developer becomes extensive. The developer can control the stochastic distribution of the objects in the scene, the scene itself, the lighting, the synthetic sensors, and the inclusion of crucial corner cases in the datasets. Eventually, the workflow also helps version and debug information for the exact reproduction of the datasets for auditing and safety.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
AI

SambaNova Systems releases enterprise-grade GPT AI-powered language model

SambaNova Systems, a company that builds advanced software, hardware, and services to run AI applications, announced the addition of the Generative Pre-trained Transformer (GPT) language model to its Dataflow-as-a-Service™ offering. This will enable greater enterprise adoption of AI, allowing organizations to launch their customized language model in much less time — less than one month, compared to nine months or a year.

“Customers face many challenges with implementing large language models, including the complexity and cost,” said R “Ray” Wang, founder and principal analyst of Constellation Research. “Leading companies seek to make AI more accessible by bringing unique large language model capabilities and automating out the need for expertise in ML models and infrastructure.”

Natural language processing

The addition of GPT to SambaNova’s Dataflow-as-a-Service increases its Natural Language Processing (NLP) capabilities for the production and deployment of language models. This model uses deep learning to produce human-like text for leveraging large amounts of data. The extensible AI services platform is powered by DataScale®, an integrated software, and hardware system using Reconfigurable Dataflow Architecture™, as well as open standards and user interfaces.

OpenAI’s GPT-3 language model also uses deep learning to produce human-like text, much like a more advanced autocomplete program. However, its long waitlist limits the availability of this technology to a few organizations. SambaNova’s model is the first enterprise-grade AI language model designed for use in most business and text- and document-based use cases. Enterprises can use its low-code API interface to quickly, easily, and cost-effectively deploy NLP solutions at scale.

“Enterprises are insistent about exploring AI usage for text and language purposes, but up until now it hasn’t been accessible or easy to deploy at scale,” said Rodrigo Liang, CEO, and cofounder of SambaNova. “By offering GPT models as a subscription service, we are simplifying the process and broadening accessibility to the industry’s most advanced language models in a fraction of the time. We are arming businesses to compete with the early adopters of AI.”

GPT use cases

There are several business use cases for Dataflow-as-a-Service equipped with GPT, including sentiment analysis, such as customer support and feedback, brand monitoring, and reputation management. This technology can also be used for document classification, such as sorting articles or texts and routing them to relevant teams, named entity recognition and relation extraction in invoice automation, identification of patient information and prescriptions, and extraction of information from financial documents.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link

Categories
AI

OpenAI releases Triton, a programming language for AI workload optimization

All the sessions from Transform 2021 are available on-demand now. Watch now.


OpenAI today released Triton, an open source, Python-like programming language that enables researchers to write highly efficient GPU code for AI workloads. Triton makes it possible to reach peak hardware performance with relatively little effort, OpenAI claims, producing code on par with what an expert could achieve in as few as 25 lines.

Deep neural networks have emerged as an important type of AI model, capable of achieving state-of-the-art performance across natural language processing, computer vision, and other domains. The strength of these models lies in their hierarchical structure, which generates a large amount of highly parallelizable work well-suited for multicore hardware like GPUs. Frameworks for general-purpose GPU computing such as CUDA and OpenCL have made the development of high-performance programs easier in recent years. Yet, GPUs remain especially challenging to optimize, in part because their architectures rapidly evolve.

Domain-specific languages and compilers have emerged to address the problem, but these systems tend to be less flexible and slower than the best handwritten compute kernels available in libraries like cuBLAS, cuDNN or TensorRT. Reasoning about all these factors can be challenging even for seasoned programmers. The purpose of Triton, then, is to automate these optimizations, so that developers can focus on the high-level logic of their code.

“Novel research ideas in the field of deep learning are generally implemented using a combination of native framework operators … [W]riting specialized GPU kernels [can improve performance,] but [is often] surprisingly difficult due to the many intricacies of GPU programming. And although a variety of systems have recently emerged to make this process easier, we have found them to be either too verbose, lack flexibility, generate code noticeably slower than our hand-tuned baselines,” Philippe Tillet, Triton’s original creator, who now works at OpenAI as a member of the technical staff, wrote in a blog post. “Our researchers have already used [Triton] to produce kernels that are up to 2 times more efficient than equivalent Torch implementations, and we’re excited to work with the community to make GPU programming more accessible to everyone.”

Simplifying code

According to OpenAI, Triton — which has its origins in a 2019 paper submitted to the International Workshop on Machine Learning and Programming Languages — simplifies the development of specialized kernels that can be much faster than those in general-purpose libraries. Its compiler simiplifies code and automatically optimizes and parallelizes it, converting it into code for execution on recent Nvidia GPUs. (CPUs and AMD GPUs and platforms other than Linux aren’t currently supported.)

“The main challenge posed by our proposed paradigm is that of work scheduling — i.e., how the work done by each program instance should be partitioned for efficient execution on modern GPUs,” Tillet explains in Triton’s documentation website. “To address this issue, the Triton compiler makes heavy use of block-level data-flow analysis, a technique for scheduling iteration blocks statically based on the control- and data-flow structure of the target program. The resulting system actually works surprisingly well: our compiler manages to apply a broad range of interesting optimization automatically.”

The first stable version of Triton, along with tutorials, is available from the project’s GitHub repository.

VentureBeat

VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.

Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the subjects of interest to you
  • our newsletters
  • gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
  • networking features, and more

Become a member

Repost: Original Source and Author Link