Crafting the Next Generation of AI
Artificial intelligence lies at the core of the fourth Industrial Revolution.
Our goal at Deci is to enable more deep learning models to fully perform in production and fulfill their true potential.
We took an innovative approach, using AI itself to craft the next generation of deep learning.
A multi-talented team with a shared passion for technology
Founded in 2019 by world recognized experts in AI, with an innate passion for creative innovation, we forged a talented team of deep learning researchers and engineers.
Deci’s team consists of top AI researcher with advanced academic degrees from prestigious universities, as well as exceptional backgrounds in elite units and leading tech companies.
Our experts are fully dedicated to advancing the state-of-the-art and have co-authored dozens of publications in
top-tier machine learning venues such as NeurIPS, ICML, ICLR, CVPR, and ACL.
Founding date
Total funding
Improvements in AI
models’ performance
Growing ecosystem of
customers, DL
community, partners
Backend Engineer
Deep Learning Engineer
Deep Learning Engineer
Research Scientist
Enterprise Account Executive
Deep Learning Software Engineer
Deep Learning Software Engineer
Director of Sales Development
Deep Learning Product Manager
Deep Learning Research Engineer
Deep Learning Engineer
VP Finance
Deep Learning Research Engineer
Enterprise Account Executive
Deep Learning Engineer
Co-Founder & Chief Scientist
Co-Founder & COO
Content Marketing Manager
VP Sales
Account Executive
Senior Frontend Engineer
Sales Development Representative
Deep Learning Research Engineer
Sales Engineer
Deep Learning Research Engineer
Co-Founder & CEO
Financial Controller
Full Stack Developer
Sales Engineer
Sales Development Representative
Deep Learning Research Engineer
VP Strategy & Business Development
Technical Product Marketing Manager
Deep Learning Engineer
Deep Learning Engineer
Full Stack Engineer
VP People
NLP Research Engineer
Marketing Manager
Deep Learning Software Engineer
Head of AI
Project Manager
NLP Research Engineer
HR Manager
Sales Development Representative
Lead Bookeeper
AI Product Manager
IT System Admin
Deep Learning Research Engineer
Deep Learning Engineer
Deep Learning Research Engineer
Customer Facing Deep Learning Group Manager
DevOps Engineer
DevRel Manager
VP Marketing
Deep Learning Research Scientist
Deep Learning Engineer, Team Lead
Deep Learning Software Engineer
VP Operations & General Manager, US
Director of Strategic Partnership
Employee Experience & Office Manager
Research Scientist
Deep Learning Research Engineer
Technical Product Manager
Research Scientist
Deep Learning Software Engineer
VP Engineering
Deci’s new platform gives enterprises the freedom to deploy high-performance LLMs, while keeping data secure and inference costs under control
TEL AVIV, Israel – March 14, 2024 – Deci, the deep learning company harnessing artificial intelligence (AI) to build AI, today announced the launch of its Generative AI Platform, a comprehensive solution designed to meet the efficiency and privacy needs of enterprises. Deci’s Gen AI platform features a new series of proprietary, fine-tunable large language models (LLMs), an inference engine, and an AI inference cluster management solution.
The first model being released in the series, Deci-Nano, has advanced language and reasoning capabilities that make it ideal for a broad spectrum of applications, such as analyzing financial and medical documents, assisting in copywriting, acting as a chatbot, summarizing, and brainstorming.
Deci-Nano sets itself apart from other models through its exceptional computational and memory efficiency, further amplified by Deci’s inference engine, Infery LLM, for unparalleled cost-efficiency and real-time latency. Compared to models with similar abilities, including Google’s Gemma 7B and Mistral 7B, Deci-Nano offers the best balance between price, quality and speed. It boasts a higher score on MT Bench, one of the most widely recognized and used LLM evaluation benchmarks. Furthermore, Deci-Nano generates text 38% faster than Mistral 7B-Instruct-v0.2 with a 60% reduction in price per 1M tokens.
Deci-Nano builds on Deci’s track record of innovative and efficient Generative AI models, including the open-source DeciLM-7B, DeciCoder 6B, DeciCoder 1B, and DeciDiffusion 2.0, which are also available through the platform. Similar to its other models, Deci-Nano was generated with Deci’s cutting-edge Automated Neural Architecture Construction (AutoNAC) engine, the most advanced Neural Architecture Search (NAS)-based technology on the market, with its focus on efficiency.
Recognizing the critical importance of data privacy, Deci’s Generative AI Platform gives customers the flexibility to deploy models through a Virtual Private Cloud (VPC) or directly within their data centers or access them through the platform’s API. This approach ensures that enterprises retain full control over their sensitive information.
“In developing our new generative AI platform, we engaged closely with the market and our enterprise customers, listening intently to their needs to understand and address the critical tradeoffs businesses make today in integrating AI models into their operations,” said Yonatan Geifman, CEO & Co-founder of Deci. “Requiring both high performance, greater control, and cost efficiency, we delivered a platform that empowers businesses with the tools they need to integrate AI safely and cost-effectively”.
With this technology, Deci AI is setting new benchmarks for operational efficiency and cost savings for enterprises.
The platform is currently available and accessible through Deci’s website.
This announcement was originally published on Cision PRWeb.
Deci’s groundbreaking new models streamline the deployment of advanced Generative Artificial Intelligence on the Qualcomm Cloud AI 100 solution, unlocking cost effective, real-time AI processing
TEL AVIV, Israel, January 17, 2024 — Deci, the deep learning company harnessing artificial intelligence (AI) to build AI, announced today it is collaborating with Qualcomm Technologies, Inc. to introduce advanced Generative Artificial Intelligence (AI) models tailored for the Qualcomm® Cloud AI 100, Qualcomm Technologies’ performance and cost-optimized AI inference solution designed for Generative AI and large language models (LLMs). This working relationship between the two companies is designed to make AI accessible for a wider range of AI-powered applications, resulting in the democratization of Generative AI’s transformative power for developers everywhere.
“Together with Qualcomm Technologies we are pushing the boundaries of what’s possible in AI efficiency and performance” said Yonatan Geifman, CEO and co-founder of Deci. “Our joint efforts streamline the deployment of advanced AI models on Qualcomm Technologies’ hardware, making AI more accessible and cost-effective, and economically viable for a wider range of applications. Our work together is a testament to our vision of making the transformational power of generative AI available to all.”
Through the relationship, Deci will work with Qualcomm Technologies to launch two groundbreaking models. The first model is DeciCoder-6B, a 6 billion parameter model for code generation engineered with a focus on performance at scale. Supporting eight programming languages (C, C#, C++, GO, RAST, Python, Java, JavaScript), it outperforms established models such as CodeGen2.5-7B, StarCoder-7B, and CodeLlama-7B. In fact, in Python, DeciCoder achieves a 3-point lead over models more than twice its size, such as StarCoderBase 15.5B. The model also stands out for its remarkable memory and computational efficiency, boasting 19x higher throughput compared to similar models when running on Qualcomm’s Cloud AI 100.
The second model, DeciDiffusion 2.0, is a 732 million parameter text-to-image diffusion model that sets new standards by outperforming Stable Diffusion v1.5, operating at 2.6 times the speed with on-par image quality. Both models are meticulously optimized to leverage the full potential of the Qualcomm Cloud AI 100 solution. These models are designed to enable users across various industries to experience exceptional performance from the outset at a more competitive price point.
Both DeciCoder-6B and DeciDiffusion 2.0 were developed using Deci’s Neural Architecture Search Technology, AutoNAC™, its proprietary, hardware-aware technology that democratizes the use of Neural Architecture Search for enterprises of all sizes. The distinctive architecture of both models ensures efficient scaling of batching while maintaining minimal memory usage and avoiding any increase in latency. Additionally, the models were designed to handle large batches, enabling maximal utilization of the computational power of the Qualcomm’s Cloud AI 100 cores. DeciCoder-6B and DeciDiffusion have been released under Apache-2.0 and CreativeML Open RAIL++-M Licenses, respectively.
This announcement was originally published on Cision PRWeb.
DeciLM-7B sets new performance standards in the large language model (LLM) space, outperforming notable open-source models such as Llama2 7B and Mistral 7B.
TEL AVIV, Israel, December 12, 2023 — Deci, the deep learning company harnessing AI to build AI, today unveiled the latest addition to its suite of innovative generative AI models, DeciLM-7B, a 7 billion parameter large language model. Building upon the success of its predecessor DeciLM 6B, DeciLM 7B is setting new benchmarks in the large language model (LLM) space, outperforming prominent open-source models such as Llama2 7B and Mistral 7B in both accuracy and efficiency.
DeciLM-7B stands out for its unmatched performance, surpassing open-source language models up to 13 billion parameters in both accuracy and speed with less computational demand. It achieves a 1.83x and 2.39x increase in throughput over Mistral 7B and Llama 2 7B, respectively, which means significantly faster processing speeds compared to competing models. Its compact design is ideal for cost-effective GPUs, striking an unparalleled balance between affordability and high-end performance.
The remarkable performance of DeciLM-7B can be further accelerated when used in tandem with Infery-LLM, the world’s fastest inference engine, designed to deliver high throughput, low latency and cost effective inference on widely available GPUs. This powerful duo sets a new standard in throughput performance, achieving speeds 4.4 times greater than Mistral 7B with vLLM without sacrificing quality. Leveraging DeciLM-7B in conjunction with Infery-LLM enables teams to drastically reduce their LLM compute expenses, while simultaneously benefiting from quicker inference times. This integration facilitates the efficient scaling of Generative AI workloads and supports the transition to more cost-effective hardware solutions.
“With the increasing use of Generative AI in various business sectors, there’s a growing demand for models that are not only highly performant but also operationally cost efficient. Our latest innovation, DeciLM-7B, combined with Infery-LLM, is a game-changer in this regard. It’s adaptable to diverse settings, including on-premise solutions, and its exceptional inference efficiency makes high-quality large language models more accessible to a wider range of users.”
Yonatan Geifman, CEO and co-founder, Deci
This synergy enables the efficient serving of multiple clients simultaneously without excessive compute costs or latency issues. This is especially crucial in sectors such as telecommunications, online retail, and cloud services, where the ability to respond to a massive influx of concurrent customer inquiries in real time can significantly enhance user experience and operational efficiency.
Licensed under Apache 2.0, DeciLM-7B is available for use and deployment anywhere, including local setups, enabling teams to fine tune for specific industry applications without compromising on data security or privacy. Its versatility allows teams to easily tailor it for unique use cases across a wide range of business applications, including content creation, translation, conversation modeling, data categorization, summarization, sentiment analysis and chatbot development, among others. When fine tuned for specific data sets, DeciLM-7B can deliver similar quality to that of much larger models such as GPT 3.5 at approximately 97% lower cost and better speed.
“With the increasing use of Generative AI in various business sectors, there’s a growing demand for models that are not only highly performant but also operationally cost efficient,” stated Yonatan Geifman, CEO and co-founder of Deci. “Our latest innovation, DeciLM-7B, combined with Infery-LLM, is a game-changer in this regard. It’s adaptable to diverse settings, including on-premise solutions, and its exceptional inference efficiency makes high-quality large language models more accessible to a wider range of users.”
DeciLM-7B’s cost-effectiveness and reduced computational demand make advanced AI technologies more accessible to businesses of all sizes, fostering innovation and driving forward the digital transformation across various sectors. With DeciLM-7B, companies can now leverage the full potential of AI without the prohibitive costs or complexities previously associated with high-end language models.
Deci AI’s introduction of DeciLM-7B builds on its track record of innovative and efficient Generative AI models, including DeciLM 6B, DeciCoder 1B, and DeciDiffusion 1.0. Similar to its other models, DeciLM 7B was generated with Deci’s cutting-edge Automated Neural Architecture Construction (AutoNAC) engine, the most advanced Neural Architecture Search (NAS)-based technology on the market, with its focus on efficiency.
This announcement was originally published on Cision PRWeb.
This pose estimation model skillfully and effectively detects individual movements while simultaneously estimating their poses, making it ideal for real-time applications on edge devices across industries.
TEL AVIV, Israel, November 7, 2023 — Deci, the deep learning company harnessing AI to build AI, announced today the launch of YOLO-NAS Pose, a groundbreaking pose estimation model generated with Deci’s cutting-edge Automated Neural Architecture Construction (AutoNAC) engine, the most advanced Neural Architecture Search (NAS)-based technology on the market. This revolutionary model is redefining capabilities in the technical pose estimation domain, demonstrating unparalleled accuracy and latency performance.
Pose estimation, a computer vision technique enabling the precise determination of human or object positions in space, is a broad spectrum of sectors, including monitoring patient movements in healthcare, evaluating athletic performances, and crafting intuitive human-computer interfaces to enhance the capabilities of robotic systems. With superior latency-accuracy performance that eclipses other state-of-the-art models in the space, including YOLOv8 Pose, YOLO-NAS Pose holds the potential to transform industries, including healthcare, security, and beyond.
With four bespoke size variants (Nano, Small, Medium & Large), YOLO-NAS Pose addresses a wide range of computational needs. All four variants deliver significantly higher accuracy with similar or lower latency compared to their YOLOv8 Pose equivalent model variants. When comparing across variants, a significant boost in speed is evident. For example, the YOLO-NAS Pose M variant boasts 38% lower latency and achieves a +0.27 AP higher accuracy over YOLOV8 Pose L, measured on Intel Gen 4 Xeon CPUs. YOLO-NAS Pose excels at efficiently detecting objects while concurrently estimating their poses, making it the go-to solution for applications requiring real time insights.
“We’re excited to introduce YOLO-NAS Pose, a testament to our commitment to pushing the boundaries of AI for practical, real-world applications. Our AutoNAC technology is the powerhouse behind this advancement, allowing us to consistently craft models that not only achieve new performance heights but also unlock transformative use cases. We believe this model will be a vital asset to developers in the field of pose estimation and look forward to seeing the innovative ways in which it will be employed.”
Yonatan Geifman, CEO and Co-founder of Deci
YOLO-NAS Pose model architecture is available under an open source license. The pre-trained weights are available for non commercial use.
Deci’s new model follows the trailblazing success of YOLO-NAS, an object detection model that garnered widespread acclaim earlier this year. Released earlier this year, the YOLO-NAS model took the developer community by storm, with fantastic applications of the technology shared across the board. Alongside YOLO-NAS, Deci’s AutoNAC has generated some of the world’s most efficient computer vision and Generative AI models such as DeciCoder, DeciLM 6B, DeciDiffusion, and many others. In the case of YOLO-NAS Pose, AutoNAC created this foundation model featuring an innovative head design, meticulously optimized for dual objectives: locating individuals and estimating their poses.
For those keen on delving deeper into YOLO-NAS Pose or interested in requesting a demo, go to https://deci.ai/
This announcement was originally published on Cision PRWeb.
The DeciDiffusion and DeciLM models, as well as Infery LLM Software Development Kit (SDK), accelerate enterprise teams’ journey towards implementing cost-effective generative AI solutions.
TEL AVIV, Israel, September 13, 2023 — Deci, the deep learning company harnessing AI to build AI, announced today the launch of innovative generative AI Foundation Models, DeciDiffusion 1.0 and DeciLM 6B, as well as its inference Software Development Kit (SDK) – Infery LLM. These groundbreaking releases are setting a new benchmark for performance and cost efficiency in the realm of generative AI.
The intensive computational requirements for training and inference of generative AI models hinder teams from cost-effectively launching and scaling gen AI applications. Deci’s innovations directly address this gap, making scaling inference efficient, cost-effective, and ready for enterprise-grade integration. By using Deci’s open-source generative models and Infery LLM, AI teams can reduce their inference compute costs by up to 80% and use widely available and cost-friendly GPUs such as the NVIDIA A10 while also improving the quality of their offering. The models introduced by Deci cater to diverse applications, ranging from content and code generation to image creation and chat applications, among many others.
Models introduced by Deci include ‘DeciDiffusion 1.0’, a blazing-fast text-to-image model that generates quality images in less than a second, 3 times faster than the renowned Stable Diffusion 1.5 model. Next in the spotlight is DeciLM 6B, a 5.7 billion parameter model. While its accuracy stands toe-to-toe with industry giants like LLaMA 2 7B, Falcon-7B, and MPT-7B, what truly sets it apart is its blazing inference speed—clocking in at an astonishing 15 times faster than the Meta LLaMA 2 7B. Rounding out the lineup is ‘DeciCoder,’ a 1 billion parameter code generation LLM released a few weeks ago. Not only do these models deliver unparalleled inference speed, but they also provide equivalent or better accuracy.
“For generative AI to truly revolutionize industries, teams need mastery over model quality, the inference process, and the ever-pivotal cost factor,” said Yonatan Geifman, CEO and co-founder of Deci. “At Deci, our journey and extensive collaborations with the world’s AI elite have equipped us to craft a solution that’s nothing short of transformative for enterprises diving into Generative AI. With our robust array of open-source models and cutting-edge tools, we’re setting the stage for teams to redefine excellence in their generative AI ventures.”
These models, crafted using Deci’s proprietary Neural Architecture Search (AutoNAC™) technology, stand as some of the most efficient and effective generative AI models in today’s market. Alongside its Foundation Models, Deci introduces Infery LLM – an inference SDK that enables developers to gain a significant performance speed-up on existing LLMs while retaining the desired accuracy. Never-before-seen inference efficiency emerges from combining Deci’s open-source models and Infery LLM. This remarkable performance is powered by unique features such as continuous batching, advanced selective quantization, and ultra-efficient beam search, among others.
Another key differentiator of Deci’s offerings is accessibility. Unlike closed-source API models, Deci provides unrestricted access to models that can be self-hosted anywhere. This not only ensures superior performance and significantly reduced inference costs when scaled but also grants users more customization options and reduces concerns over data privacy and compliance.
“In the current landscape, where the importance of AI continues to expand, the distinction between evolution and extinction lies in the rapid, cost-effective implementation of AI tools,” said Prof. Ran El Yaniv, Chief Scientist and co-founder of Deci. “With Deci’s groundbreaking solutions, companies receive both enterprise-grade quality and control, as well as the flexibility to customize models and the inference process according to their precise requirements. This commitment ensures unmatched excellence and a lasting competitive edge.”To see the full suite of Deci generative AI offerings, visit Deci’s website.
This announcement was originally published on Cision PRWeb.
This innovative leap in code generation, powered by a 1B-parameter LLM, sets the bar for unprecedented efficiency and performance
TEL AVIV, Israel, August 16, 2023 — Deci, the deep learning company harnessing AI to build AI, today released DeciCoder, its inaugural foundation model in generative AI helping users generate programming language code. This groundbreaking Large Language Model (LLM), dedicated to code generation with 1 billion parameters and an expansive 2048-context window, surpasses results released in equivalent models and effectively redefines the standards of efficient code generation.
DeciCoder’s unmatched throughput and low memory footprint enables teams to achieve extensive code generation with low latency, and migrate workloads to more affordable and widely available GPUs such as the NVIDIA A10G, resulting in substantial cost savings. When DeciCoder was benchmarked on Hugging Face Inference Endpoints against well-established code LLMs such as SantaCoder, DeciCoder showcased a 22% increase in throughput, a significant reduction in memory usage, and a 1.5-2.4 percentage point improvement in accuracy on the HumanEval benchmark. Notably, when combining DeciCoder with Deci’s LLM inference acceleration library, Infery, its throughput outperforms that of SantaCoder by 350%.
“From enabling the fastest and most cost-efficient deployment for enterprises, to now branching into generative AI, we are relentlessly pushing boundaries and empowering developers with the advanced models and tools needed to effectively implement AI-powered applications across industries,” said Yonatan Geifman PhD, CEO & co-founder of Deci. “Utilizing DeciCoder means less operations during inference, which translates to lower computational costs.”
DeciCoder was generated using Deci’s proprietary Automated Neural Architecture Construction (AutoNAC) engine, the most advanced Neural Architecture Search (NAS)-based technology on the market. AutoNAC identifies the ideal architecture that strikes a perfect balance between accuracy and processing speed, tailored for distinct data features, tasks, performance goals, and inference environment. Deci’s AutoNAC has generated some of the world’s most efficient computer vision and NLP models such as YOLO-NAS, DeciBERT and DeciSeg, among others.
The rollout of DeciCoder is the first in a series of the highly anticipated releases outlining Deci’s Generative AI offering, which are due to be released in the coming weeks. DeciCoder and its pre-trained weights are available under the permissive Apache 2.0 License, granting developers broad usage rights and positioning the model for real-world, commercial applications.
To gain early access to Deci’s Infery and upcoming generative models, please visit https://deci.ai/get-early-access-deci-generative-ai/
This announcement was originally published on Cision PRWeb.
DataGradients enables data scientists to avoid pitfalls and save precious development time with better visibility into their data, streamlining the model design and training processes
TEL AVIV, Israel, July 12, 2023 — Deci, the deep learning company harnessing AI to build AI, today announced the release of DataGradients, a free, open-source tool for profiling computer vision datasets and distilling critical insights.
In the realm of computer vision, an AI model’s power is inherently linked to the quality of the data used for training. Identifying issues in the dataset is paramount, as it not only helps practitioners steer clear of training roadblocks but also sheds light on a model’s potential underperformance. A good read on a dataset’s attributes can help streamline decisions like the appropriate model choice, the best loss function, and the ideal optimization method.
“DataGradients empowers data scientists to streamline their model development and training process, with crystal-clear visibility into their data. With DataGradients, we’ve made it that much easier to extract actionable insights from one’s datasets,” said Yonatan Geifman, CEO and co-founder of Deci. “DataGradients marks our third tool released as open source to the benefit of the wider AI community, following our launch of SuperGradients, our free, open-source training library for PyTorch-based deep learning models, and YOLO-NAS, our groundbreaking object detection foundation model.”
Using DataGradients, data scientists can easily analyze the health of their data with one line of code, swiftly identifying problems such as corrupted data, distributional shifts between train and test sets, duplicate annotations, among many others. Users then receive actionable insights on how to proactively mitigate such issues to streamline their model design and training processes, thereby ensuring optimal performance and reliable results.
If you would like to profile your data or start training your models, visit DataGradients on Deci’s GitHub repository.
This announcement was originally published on Cision PRWeb.
Deci awarded top honors by a panel of independent industry experts
TEL AVIV, Israel, May 23, 2023 — Deci, the deep learning company harnessing AI to build AI, today announced that its deep learning development platform was selected as winner of the 2023 Edge AI and Vision Product of the Year Awards for “Best Edge AI Developer Tool.” The awards recognize the innovation and excellence of the industry’s leading technology companies that are enabling practical visual AI and computer vision.
Edge AI applications are becoming increasingly common, yet AI teams are still facing challenges when trying to reach sufficient inference performance and in some cases are not able to deploy or cost-efficiently scale their models on the target edge hardware.
“Deci is honored to be recognized as the winner of 2023 Best Edge AI Developer Tool of the Year,” said Yonatan Geifman, CEO and co-founder of Deci. “Our mission at Deci is to empower AI teams with tools to eliminate development bottlenecks and reach efficient inference performance at a faster rate. We are proud to be serving leading enterprises which deploy on edge devices across verticals and allowing them to achieve unparalleled inference performance, while also significantly reducing their development time and compute costs.”
Deci’s deep learning development platform is utilized by enterprises across industries including consumer and retail, smart cities applications, automotive, robotics, sports, smart manufacturing, and smart agriculture.
“Congratulations to Deci for earning the distinction of Best Edge AI Developer Tool for its Deep Learning Platform,” said Jeff Bier, founder of the Edge AI and Vision Alliance. “As edge AI proliferates into many industries and applications, ease of developing and deploying optimized models becomes paramount. We applaud Deci for their innovative approach to address this challenge.”
“Congratulations to Deci for earning the distinction of Best Edge AI Developer Tool for its Deep Learning Platform. As edge AI proliferates into many industries and applications, ease of developing and deploying optimized models becomes paramount. We applaud Deci for their innovative approach to address this challenge.”
Jeff Bier, founder of the Edge AI and Vision Alliance
Powered by Neural Architecture Search (NAS), the Deci platform offers advanced tools to build, train, optimize, and deploy highly accurate and efficient models to any environment, including mobile, laptops, and other edge devices, as well as the cloud and data centers. Using Deci’s NAS engine, called AutoNAC, data scientists and machine learning engineers are able to develop powerful models tailored to their specific hardware and use case and achieve outstanding performance, even on resource constrained edge devices. AutoNAC-generated models outperform well known state of the art models by 3-10x and deliver an optimal balance between accuracy and latency (or throughput).
Leading companies use Deci’s deep learning platform to accelerate inference performance, enable new use cases on edge devices, migrate workloads from cloud to edge, reduce their cloud compute cost, and significantly shorten time to market.
The Edge AI and Vision Product of the Year Awards are judged by an independent, expert panel and solutions are evaluated based on innovation, impact on customers and the market, and competitive differentiation.
This announcement was originally published on Cision PRWeb.
The new YOLO-NAS delivers state-of-the-art object detection capabilities with unparalleled accuracy, outperforming competing notable models such as YOLOv6, v7 & v8.
TEL AVIV, Israel, May 3, 2023 — Deci, the deep learning company harnessing AI to build AI, today announced the release of YOLO-NAS, its new deep learning model providing superior real-time object detection capabilities and production-ready performance. This foundation model was generated by Deci’s Neural Architecture Search Technology, AutoNAC™, and delivers unparalleled accuracy and speed, outperforming competing models, most notably YOLOv6, YOLOv7 and YOLOv8.
The world of object detection has undergone significant progress over the past few years, with YOLO models leading the charge. However, limitations and challenges in existing YOLO models, such as inadequate quantization support and insufficient accuracy-latency tradeoffs, have driven the need for continuous innovation. Deci’s new YOLO-NAS addresses these concerns by pushing the boundaries of object detection with superior real-time capabilities.
“The release of YOLO-NAS is a major leap forward for inference performance and efficiency of object detection models, addressing the limitations of previous YOLO models and offering unprecedented adaptability for diverse tasks and hardware,” said Yonatan Geifman, CEO and co-founder of Deci.
Deci’s AutoNAC is a groundbreaking technology that democratizes the use of Neural Architecture Search for every organization and helps teams quickly generate custom, fast, accurate and efficient deep learning models. AutoNAC generates best-in-class deep learning model architectures for any task in any environment, delivering the best balance between accuracy and inference speed. In addition to being data and hardware aware, the AutoNAC engine considers other components in the inference stack, including compilers and quantization.
“While Deci AutoNAC generated the best general purpose YOLO version to date, we know that there is no such thing as a “one-size-fits-all” model. Trying to use the same off-the-shelf model for live video streams analysis on edge devices and for detection on cloud GPUs will result in suboptimal performance,” said Prof. Ran El-Yaniv, co-founder and Chief Scientist at Deci.
When it comes to computer vision, the conventional approach falls short when aiming to achieve top performance, and this is particularly notable in edge AI deployments. Ideal neural architectures must meticulously consider the fine details including image resolution and object size, as well as hardware attributes, such as parallelization capabilities, operator efficiencies, and memory cache size.
El-Yaniv further explains, “The devil is in the details, and designing such optimal architectures is an incredibly complex task, often too difficult for humans to tackle alone. Deci’s pioneering AutoNAC engine empowers AI teams to construct state-of-the-art architectures that impeccably align with their applications, delivering unparalleled results.”
As demonstrated in the chart below, the YOLO-NAS (m) model delivers 50% (x1.5) increase in throughput and 1 mAP better accuracy compared to other SOTA YOLO models on the NVIDIA T4 GPU.
The YOLO-NAS model is pre-trained on well-known datasets including COCO, Objects365, and Roboflow 100, making it extremely suitable for downstream Object Detection tasks in production environments. The new model is available under an open-source license with pre-trained weights available for research use (non-commercial) on SuperGradients, Deci’s PyTorch-based, open-source, computer vision training library. With SuperGradients, users can train models from scratch or fine-tune existing ones, leveraging advanced built-in training techniques like Distributed Data Parallel, Exponential Moving Average, Automatic mixed precision, and Quantization Aware Training.
Deci’s SuperGradients library incorporates the latest advancements in deep learning training. If you would like to train and deploy the YOLO-NAS or any other computer vision model, please go to our GitHub repository.
This announcement was originally published on Cision PRWeb.
Deci achieves the highest inference speed ever to be published at MLPerf for NLP, while also delivering the highest accuracy.
Tel Aviv, Israel, April 5, 2023 — Deci, the deep learning company harnessing Artificial Intelligence (AI) to build better AI, today announced results for its Natural Language Processing (NLP) model submitted to the MLPerf Inference v3.0 benchmark suite under the open submission track. Notably, the NLP model, generated by Deci’s Automated Neural Architecture Construction (AutoNAC) technology, dubbed DeciBERT-Large, delivered a record-breaking throughput performance of more than 100,000 queries per second on 8 NVIDIA A100 GPUs while also delivering improved accuracy. Also, Deci delivered unparalleled throughput performance per TeraFLOPs, outperforming competing submissions made on even stronger hardware setups.
Running successful inference at scale requires meeting various performance criteria such as latency, throughput, and model size, among others. Optimizing inference performance after a model has already been developed is an especially cumbersome and costly process, often leading to project delays and failures. Accounting for the inference environment and production constraints early in the development lifecycle can significantly reduce the time and cost of fixing potential obstacles to trying to deploy models.
“These results demonstrate once again the power of Deci’s AutoNAC technology, which is leveraged today by leading AI teams to develop superior deep learning applications, faster,” said Prof. Ran El-Yaniv, Deci’s chief scientist and co-founder. “With Deci’s platform, teams no longer need to compromise either accuracy or inference speed, and achieve the optimal balance between these conflicting factors by easily applying Deci’s advanced optimization techniques”. Deci’s model was submitted under the offline scenario in MLPerf’s open division in the BERT 99.9 category. The goal was to maximize throughput while keeping the accuracy within a 0.1% margin of error from the baseline, which is 90.874 F1 (SQUAD).
For the submission, Deci leveraged its deep learning development platform powered by its proprietary AutoNAC engine. The AutoNAC engine empowers teams to develop hardware aware model architectures tailored for reaching specific performance targets on their inference hardware. Models built and deployed with Deci typically deliver up to 10X increase in inference performance with comparable or higher accuracy relative to state of the art open source models. This increase in speed translates into a better user experience and a significant reduction in inference compute costs.
In this case, AutoNAC was used by Deci to generate model architectures tailored for various NVIDIA accelerators and presented unparalleled performance on the NVIDIA A30 GPU, NVIDIA A100 GPU (1 & 8 unit configurations), and the NVIDIA H100 GPU.
The below chart illustrates the throughput performance per TeraFLOPs as achieved by Deci and other submitters within the same category. Deci delivered the highest throughput per TeraFLOPs while also improving the accuracy. This inference efficiency translates into significant cost savings on compute power and a better user experience. Instead of relying on more expensive hardware, teams using Deci can now run inference on NVIDIA’s A100 GPU, achieving 1.7x faster throughput and +0.55 better F1 accuracy, compared to when running on NVIDIA’s H100 GPU. This means a 68%* cost savings per inference query.
Other benefits of Deci’s results include the ability to migrate from multi-gpu to a single GPU and lower inference cost and reduced engineering efforts. For example, ML engineers using Deci can achieve a higher throughput on one H100 card than on 8 NVIDIA A100 cards combined. In other words, with Deci, teams can replace 8 NVIDIA A100 cards with just one NVIDIA H100 card, while getting higher throughput and better accuracy (+0.47 F1).
On the NVIDIA A30 GPU, which is a more affordable GPU, Deci delivered accelerated throughput and a 0.4% increase in F1 accuracy compared to an FP32 baseline.
By using Deci, teams that previously needed to run on an NVIDIA A100 GPU can now migrate their workloads to the NVIDIA A30 GPU and achieve 3x better performance then they previously had for roughly a third of the compute price. This means dramatically better performance for significantly less inference cloud cost.
Hardware | Other Submitters’ Throughput | Deci’s Throughput | BERT F1 Accuracy | Deci Optimized F1 Accuracy | Accuracy Increase |
NVIDIA A30 GPU | 866 | 5,885 | 90.874 | 91.281 | 0.4076 |
NVIDIA A100 GPU, 80GB | 1,756 | 13,377 | 90.874 | 91.430 | 0.5560 |
8 x NVIDIA A100 GPU | 13,967 | 103,053 | 90.874 | 91.430 | 0.5560 |
NVIDIA H100 PCIe GPU | 7,921 | 17,584 | 90.874 | 91.346 | 0.4722 |
Recently, Deci launched a new version of its deep learning platform, supporting generative AI model optimization and continuing to help developers further simplify the AI lifecycle.
This announcement was originally published on Cision PRWeb.
Deci’s new version simplifies the AI development lifecycle and breaks down additional barriers on the journey to production.
Tel Aviv, Israel, February 22, 2023 — Deci, the deep learning company harnessing AI to build AI, today released a new version of its deep learning platform, helping developers further simplify the AI lifecycle and break down additional barriers on the journey to production. With this new version, Deci’s platform will now support generative AI model optimization, as well as enhance existing inference acceleration and optimization capabilities with no accuracy degradation.
“With the generative AI revolution upon us, it’s more important than ever for the industry to explore alternative approaches in order to make such models scalable. The hidden costs in terms of computational and financial resources required for these models to run are just too high, particularly the inference costs which are even more substantial in generative AI,” said Yonatan Geifman, Co-Founder and CEO of Deci. “The need to optimize these generative AI models has risen dramatically over the last few months. Fortunately, with our platform, developers can build, optimize, and deploy smaller, more specialized models for specific applications, effectively unlocking deep learning’s many advantages while removing development and production barriers.”
Deci’s platform empowers AI teams in developing production-grade deep learning models. The platform simplifies and accelerates the development process with advanced tools to build, train, optimize, and deploy highly accurate and efficient models to any environment. The platform is powered by Deci’s proprietary AutoNAC (Automated Neural Architecture Construction) engine, based on Neural Architecture Search (NAS) Technology, that generates best-in-class deep learning model architectures for any task in any environment. The platform enables AI teams to accelerate inference performance by up to 10X, reduce inference costs by 5X, shorten development time by 80%, and lower development costs by 30%.
Deci’s new version includes:
This announcement was originally published on Cision PRWeb.
The Intel-Deci breakthrough enables AI developers to achieve GPU-like AI inference performance on CPUs in production for both computer vision and NLP tasks
TEL AVIV, Israel, January 25, 2023 — Deci, the deep learning company building the next generation of AI, announced breakthrough performance on Intel’s newly released 4th Gen Intel® Xeon® Scalable processors, code-named Sapphire Rapids. By optimizing the AI models which run on Intel’s new hardware, Deci enables AI developers to achieve GPU-like inference performance on CPUs in production for both Computer Vision and Natural Language Processing (NLP) tasks.
Deci utilized its proprietary AutoNAC (Automated Neural Architecture Construction) technology to generate custom hardware-aware model architectures that deliver unparalleled accuracy and inference speed on the Intel Sapphire Rapids CPU. For computer vision, Deci delivered a 3.35x throughput increase, as well as a 1% accuracy boost, when compared to an INT8 version of a ResNet50 running on Intel Sapphire Rapids. For NLP, Deci delivered a 3.5x acceleration compared to the INT8 version of the BERT model on Intel Sapphire Rapids, as well as a +0.1 increase in accuray. All models were compiled and quantized to INT8 with Intel® Advanced Matrix Extensions (AMX) and Intel extension for PyTorch.
“This performance breakthrough marks another chapter in the Deci-Intel partnership which empowers AI developers to achieve unparalleled accuracy and inference performance with hardware-aware model architectures powered by NAS,” said Yonatan Geifman, CEO and Co-Founder of Deci. “We are thrilled to enable our joint customers to achieve scalable, production grade performance, within days”.
Deci and Intel have maintained broad strategic business and technology collaborations since 2019, most recently announcing the acceleration of deep learning models using Intel Chips with Deci’s AutoNAC technology . Deci is a member of the Intel Disruptor program and has collaborated with Intel on multiple MLPerf submissions. Together, the two are enabling new deep learning based applications to run at scale on Intel CPUs, while reducing development costs and time to market.
If you are using CPUs for deep learning inference or planning to do so, talk with Deci’s experts to learn how you can quickly obtain better performance and ensure maximum hardware utilization. To learn more about the Deci-Intel collaboration, go to https://deci.ai/intel/.
This announcement was originally published on Cision PRWeb.
With the power of Deci’s Automated Neural Architecture Construction (AutoNAC) technology, developers are better suited to build, optimize, and deploy more powerful deep learning models using Intel Processors
Tel Aviv, Israel, November 10, 2022 – Deci, the deep learning company harnessing Artificial Intelligence (AI) to build AI, today announced a new strategic collaboration with Intel to accelerate the journey towards more scalable AI. By combining Deci’s proprietary Automated Neural Architecture Construction (AutoNAC) technology with Intel processor technology, the two companies will further optimize deep learning inference, enabling developers everywhere to build, optimize, and deploy more accurate, fast and efficient models for the edge, datacenter, and cloud.
As part of the Deci-Intel collaboration continues, Deci recently joined the Intel Disruptor Program, which provides technical enablement and go-to-market activities for participants. Deci-Intel collaboration was initiated by Intel Labs and Deci was one of the first companies to join Intel Ignite, an accelerator program designed to support innovative startups in advancing new technologies in disruptive markets.
Deci is now working with Intel to demonstrate AutoNAC’s performance on 4th Gen Intel Xeon Scalable processors, formerly codenamed Sapphire Rapids. Together, Deci and Intel are making significant steps towards enabling breakthrough deep learning inference on CPUs, a break from tradition as GPUs have generally been the default choice for AI tasks.
“As a result of our collaboration with Intel, we’ve seen exciting achievements in such a short period – deep learning at scale on CPUs is more feasible than ever before,” said Yonatan Geifman, CEO and Co-Founder of Deci. “We expect that our joint activities will only further propel AI accessibility, dramatically optimizing deep learning inference for any task in any environment.”
Deci and Intel first announced their broader strategic business and technology collaboration in 2021, following several groundbreaking submissions at MLPerf. In 2022, Deci announced its results for both its Computer Vision (CV) and Natural Language Processing (NLP) models that were submitted to the MLPerf v2.0 Datacenter Open division. On 2nd Gen and 3rd Gen Intel Xeon Scalable processors, Deci’s AutoNAC generated models that delivered breakthrough accuracy and throughput performance- for their CV submission, Deci delivered +1.74% improvement in accuracy and 4x improvement in throughput, while for their NLP submission, Deci improved accuracy by +1.03% and throughput performance by 5x. This was a continuation of their MLPerf results in 2021 where on several Intel Xeon Scalable Processors, Deci reduced the submitted models’ latency by a factor of up to 11.8x and increased throughput by up to 11x– all while preserving the model’s accuracy within 1%.
“The journey towards more scalable AI has never been more important as this technology continues to unlock groundbreaking use cases impacting industries across the board,” Arijit Bandyopadhyay – CTO Enterprise Analytics and AI, Head of Strategy – Cloud and Enterprise, Data Center Group, Intel Corporation. “We’ve seen firsthand the groundbreaking nature of Deci’s AutoNAC technology and its ability to automatically generate deep learning. Working together, customers will see the true value of innovative AI on Intel technologies.”
Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries.
This announcement was originally published on Cision PRWeb.
Deci’s Deep Learning NAS technology automatically generated a new family of models dubbed DeciSeg, which deliver unparalleled inference performance and accuracy
Tel Aviv, Israel, September 22, 2022 – Deci, the deep learning company harnessing AI to build AI, today announced a new set of industry-leading semantic segmentation models, dubbed DeciSeg. Deci’s proprietary Automated Neural Architecture Construction (AutoNAC) technology automatically generated semantic segmentation models that significantly outperform the most powerful models publicly available, such as the MobileViT released by Apple, and the DeepLab family released by Google. Deci’s models deliver more than 2x lower latency, as well as 3-7% higher accuracy.
Semantic segmentation is one of the most widely used computer vision tasks across many business verticals, including automotive, smart cities, healthcare, and consumer applications, and is often required for many edge AI applications. However, significant barriers exist to running semantic segmentation models directly on edge devices, such as high latency and the inability to deploy those models due to their size.
With DeciSeg models, semantic segmentation tasks that previously could not be carried out at the edge because they were too resource intensive are now possible. This allows companies to develop new use cases and applications on edge devices, reduce inference costs (since AI practitioners will no longer need to run these tasks in expensive cloud environments), open new markets, and shorten development times.
“DeciSegs are an example of the power of Deci’s AutoNAC engine capabilities to generate custom hardware-aware deep learning models with unparalleled performance on any hardware. AI teams can easily use DeciSegs models or leverage Deci’s AutoNAC engine to build and deploy custom models that run real-time computer vision tasks on their edge devices.” said Yonatan Geifman, PhD, co-founder and CEO of Deci.
Deci’s platform has a proven-track record in enabling AI at the edge and empowering AI teams to build and deploy production grade deep learning models. Earlier this year, Deci announced the discovery of DeciNets for CPUs, which reduced the gap between a model’s inference performance on a GPU versus a CPU by half, without sacrificing the model’s accuracy, enabling AI to run on lower cost, resource constrained hardware.
“In the world of automated deep neural network design and construction, Deci’s AutoNAC technology is a game changer. It uses deep learning to search vast spaces of neural networks for the model most appropriate for a particular task and particular AI chip. In this case, AutoNAC was applied to the Pascal VOC Semantic Segmentation task on NVIDIA’s Jetson Xavier NX™ chip and we are very pleased with the results.” said Ran El-Yaniv, co-founder and Chief Scientist of Deci and Professor of Computer Science at the Technion – Israel Institute of Technology.
Deci’s platform is serving customers across industries in various production environments including edge, mobile, data centers and cloud. To learn more about how leading AI teams leverage Deci’s platform to build production grade models and accelerate inference performance, visit here.
This announcement was originally published on Cision PRWeb.
DeciBERT-Large substantially improved throughput performance & accuracy while also significantly reducing model size
Tel Aviv, Israel, September 8, 2022 – Deci, the deep learning company harnessing Artificial Intelligence (AI) to build better AI, announced results for its Natural Language Processing (NLP) inference model submitted to the MLPerf Inference v2.1 benchmark suite under the open submission track. Generated by Deci’s Automated Neural Architecture Construction (AutoNAC) technology, the NLP model, dubbed DeciBERT-Large, ran on Dell-PowerEdge-R7525-2 hardware using the AMD-EPYC-7773X processor. The resulting model outperformed both the throughput performance of the BERT-Large model by 6.46x and achieved a 1% boost in accuracy.
The model was submitted under the offline scenario in MLPerf’s open division in the BERT 99.9 category. The goal was to maximize throughput while keeping the accuracy within a 0.1% margin of error from the baseline, which is 90.874 F1 (SQUAD). The DeciBERT-Large model far exceeded these goals, reaching a throughput of 116 QueriesPer Second (QPS) and an F1 score of 91.08 for accuracy.
“While the key optimization objective when generating the DeciBERT model was to optimize throughput, AutoNAC also managed to significantly reduce the model size – an important accomplishment with a number of benefits including the ability to run multiple models on the same server and better utilize cache memory.” – Prof. Ran El-Yaniv, Deci’s chief scientist and co-founder of Deci
For the submission, Deci leveraged its proprietary automated Neural Architecture Construction technology (AutoNAC) engine to generate a new model architecture tailored for the AMD processor. AutoNAC, an algorithmic optimization engine generating best-in-class deep learning model architectures for any task, data set, and inference hardware, typically powers up to a 5X increase in inference performance with comparable or higher accuracy relative to state-of-the-art neural models.
“While the key optimization objective when generating the DeciBERT model was to optimize throughput, AutoNAC also managed to significantly reduce the model size – an important accomplishment with a number of benefits including the ability to run multiple models on the same server and better utilize cache memory,” said Prof. Ran El-Yaniv, Deci’s chief scientist and co-founder. “These results confirm once again the exceptional performance of our AutoNAC technology, which is applicable to nearly any deep learning domain and inference hardware”.
MLPerf gathers expert deep learning leaders to build fair and useful benchmarks for measuring training and inference performance of ML hardware, software, and services.
Deci’s NLP inference acceleration directly translates into cloud cost reduction as it enables more processes to run on the same machine in less time or alternatively it enables teams to use a more cost efficient machine while retaining the same throughput performance. For some NLP applications such as question answering, higher throughput also means better user experience as the queries are processed faster and insights can be generated in real time.
This announcement was originally published on Cision PRWeb.
Tel Aviv, Israel, August 4, 2022 – Deci, a deep learning company harnessing AI to build AI, today announced it has joined NVIDIA Metropolis — a partner program, application framework and set of developer tools that bring to market a new generation of applications and solutions to make the world’s most important spaces and operations safer and more efficient with advancements in AI vision.
Deci enables AI developers to build, optimize and deploy best-in-class deep learning models, delivering high accuracy tailored for any dataset, inference hardware, speed and size requirements. Its platform enables unparalleled inference performance on the NVIDIA Jetson edge AI platform — which includes the Jetson Orin, AGX Xavier, Xavier NX and Nano modules — as well as server-based NVIDIA GPUs. With Deci, vendors can deploy complex models onto smaller edge devices, thus achieving real-time latency and maximizing hardware utilization.
NVIDIA Metropolis makes it easier and more cost effective for enterprises, governments and integration partners to use world-class AI-enabled solutions to improve critical operational efficiency and safety problems. The NVIDIA Metropolis ecosystem contains a large and growing breadth of members who are investing in the most advanced AI techniques and most efficient deployment platforms, while using an enterprise-class approach to their solutions. Members have the opportunity to gain early access to NVIDIA platform updates to further enhance and accelerate their AI application development efforts. Further, the program offers the opportunity for members to collaborate with industry-leading experts and other AI-driven organizations.
“We are honored to be part of NVIDIA Metropolis and confident that our participation will enable us to reach more customers to support them in successfully deploying world-changing AI solutions,” said Yonatan Geifman, CEO and co-founder of Deci. “AI teams can rely on Deci’s platform to build and optimize top-notch models at the edge, a real game changer for enterprises seeking to innovate.”
This announcement was originally published on Cision PRWeb.
Deci’s deep learning development platform bridges the AI efficiency gap, empowering AI teams to efficiently build next generation deep learning applications.
Tel Aviv, Israel, July 13, 2022 – Deci, the deep learning company harnessing AI to solve the AI efficiency gap, today announced it has raised $25 million in a Series B funding round led by global software investor Insight Partners, with participation from existing investors Square Peg, Emerge, Jibe Ventures, and Fort Ross Ventures, as well as new investor ICON. The investment comes just seven months after Deci secured $21 million in Series A funding, also led by Insight Partners, bringing Deci’s total funding to $55.1 million. The funds will be used to expand Deci’s go-to-market activities, as well as further accelerate the company’s R&D efforts.
Deep learning-powered advancements in AI have led to innovations that have the potential to revolutionize services, products, and consumer applications across industries such as medicine, manufacturing, transportation, communication, and retail. However, the AI efficiency gap – a situation in which hardware is unable to meet the increasing computing demands of models that are growing in size and complexity – has proven to be an obstacle to more widespread AI commercialization. This efficiency gap means that inference is still generally bound to the cloud, where compute hardware is abundant but costs are high and concerns around data privacy and safety are prevalent.
“Deci’s deep learning development platform has a proven record of enabling companies of all sizes to do just that by providing them with the tools they need to successfully develop and deploy world-changing AI solutions – no matter the level of complexity or production environment. This funding is a vote of confidence in our work to make AI more accessible and scalable for all.”
Yonatan Geifman, CEO and co-founder of Deci
Deci’s deep learning platform helps data scientists eliminate the AI efficiency gap by adopting a more productive development paradigm. With the platform, AI developers can leverage hardware-aware Neural Architecture Search (NAS) to quickly build highly optimized deep learning models that are designed to meet specific production goals.
“The growing AI efficiency gap only further highlights the importance of ‘shifting left’ – accounting for production considerations early in the development lifecycle, which can then significantly reduce the time and cost spent on fixing potential obstacles when deploying models in production,” said Yonatan Geifman, CEO and co-founder of Deci. “Deci’s deep learning development platform has a proven record of enabling companies of all sizes to do just that by providing them with the tools they need to successfully develop and deploy world-changing AI solutions – no matter the level of complexity or production environment. This funding is a vote of confidence in our work to make AI more accessible and scalable for all.”
The platform empowers data scientists to deliver superior performance at a much lower operational cost (up to 80% reduction), reduce time to market from months to weeks and easily enables new applications on resource-constrained hardware such as mobile phones, laptops, and other edge devices.
Deci’s deep learning development platform is powered by Deci’s proprietary AutoNAC (Automated Neural Architecture Construction) technology, an algorithmic optimization engine that empowers data scientists to build best-in-class deep learning models that are tailored for any task, data set and target inference hardware. Deci’s AutoNAC engine democratizes NAS technology, something that until very recently was confined to academia or industry giants like Google due to its high cost.
“Deci’s powerful technology lets you input your AI models, data, and target hardware — whether that hardware is on the edge or in the cloud — and guides you in finding alternative models that will generate similar predictive accuracy with massively improved efficiency. We are very excited to double down on our investment in Deci, backing Yonatan and the team as they bring this critical technology to
Lonne Jaffe, Managing Director at Insight Partners
“Having a more efficient infrastructure for AI systems can make AI products qualitatively different and better, not only cheaper and faster to run,” said Lonne Jaffe, Managing Director at Insight Partners and board member at Deci. “Deci’s powerful technology lets you input your AI models, data, and target hardware — whether that hardware is on the edge or in the cloud — and guides you in finding alternative models that will generate similar predictive accuracy with massively improved efficiency. We are very excited to double down on our investment in Deci, backing Yonatan and the team as they bring this critical technology to AI builders across the world.”
Deci recently announced the launch of version 2.0 of its platform, which helps enterprises build, optimize, and deploy state-of-the-art computer vision models on any hardware and environment, including cloud, edge and mobile, with outstanding accuracy and runtime performance. Deci also announced the impressive results of its AutoNAC-generated DeciBERT models at MLPerf v2.0. For natural language processing (NLP), Deci’s models accelerated question-answering tasks’ throughput performance on various Intel CPUs by 5x (depending on the hardware type and quantization level) while also improving the accuracy by +1.03%.
Deci collaborates with various hardware manufacturers, Computer OEMs and other ML ecosystem leaders, and is an official partner of Intel, Amazon Web Services (AWS), Hewlett Packard Enterprise (HPE), and NVIDIA among others.
This announcement was originally published on Cision PRWeb.
Deci’s platform enables AI developers to build, optimize and deploy highly accurate, fast and efficient computer vision models on any hardware
Tel Aviv, Israel, May 11, 2022 – Deci, the deep learning company harnessing AI to build AI, today launched Version 2.0 of its deep learning development platform, making it easier than ever before for AI developers to build, optimize, and deploy computer vision models on any hardware and environment including cloud, edge and mobile with outstanding accuracy and runtime performance.
AI developers face an uphill struggle developing production-ready deep learning models for deployment. These challenges can be largely attributed to the AI efficiency gap facing the industry in which algorithms are growing more powerful and complex, but available compute power is not keeping pace. This gap also creates financial barriers by making the deep learning development and processing more cumbersome and expensive.
While Neural Architecture Search (NAS) has been presented as a potential solution to automate the design of superior artificial neural networks that can outperform manually-designed architectures, the resource requirements to operate such technology is excessive. To date, NAS has only been successfully implemented by tech giants like Google, Microsoft and in the confines of academia, proving its impracticality for the vast majority of developers.
In order to solve this problem, Deci’s platform, powered by its proprietary NAS engine called AutoNAC (Automated Neural Architecture Construction), enables AI developers to automatically and affordably build efficient computer vision models that deliver the highest accuracy for any given inference hardware, speed, size and targets. Models generated by Deci outperform other known state-of-the-art (SOTA) architectures by a factor of 3x-10x.
The new version of Deci’s deep learning platform makes hardware-aware NAS technology accessible to AI teams of any size, helping them eliminate complexities and focus on what they do best – build innovative computer vision applications.
Developers can start their projects with pre-trained and optimized models (DeciNets) that were generated by the AutoNAC engine for a wide range of hardware and computer vision tasks or use the AutoNAC engine to generate more custom architectures that are tailored for their specific use-cases. In addition, the platform supports teams with a wide range of tools required to develop deep learning-based applications including a hardware-aware model zoo to easily select and benchmark models and hardware, SuperGradients – an open source training library with proven recipes for faster training, automated runtime optimizations, model packaging and more.
By using Deci’s platform, AI developers achieve improved inference performance and efficiency to enable deployment on resource constrained edge devices, maximize hardware utilization and reduce training and inference cost. The entire development cycle is shortened and the uncertainty of how the model will deploy on the inference hardware is eliminated.
“The new version of Deci’s deep learning platform makes hardware-aware NAS technology accessible to AI teams of any size, helping them eliminate complexities and focus on what they do best – build innovative computer vision applications.” said Yonatan Geifman, co-founder and CEO of Deci. “We take pride in the fact that the deep learning models generated by Deci’s platform are powering AI-based applications of some of the leading enterprises worldwide. We are excited to unleash this powerful engine to help make computer vision even more widely available. Only then can we truly achieve a world where AI advances humanity without limitations, finally making AI affordable, accessible and scalable for all.”
With Deci’s Version 2.0 platform, AI developers can:
With Deci’s hardware-aware model zoo, developers can quickly measure inference time of pre-trained and optimized models on and various hardware including edge devices via Deci’s SaaS platform. Simplify the hardware and model selection process by eliminating the need to manually setup and test various combinations of models and hardware.
Automatically find accurate & efficient architectures tailored for the application, hardware and performance targets with Deci’s AutoNAC engine.
Leverage proven hyperparameter recipes and with Deci’s PyTorch based open source training library called SuperGradients.
Automatically compile and quantize your models and evaluate different production settings.
Developers can deploy their deep learning workloads on any environment with the Deci’s python based inference engine.
Deci’s platform includes three tiers:
Users can request a free trial to Deci’s community version to get started. For more information, visit us here.
This announcement was originally published on Cision PRWeb.
Deci’s submissions deliver breakthrough accuracy & throughput performance across CV and NLP models on Intel’s CPUs
Tel Aviv, Israel, April 6, 2022 – Deci, the deep learning company harnessing Artificial Intelligence (AI) to build AI, announced its results for both Computer Vision (CV) and Natural Language Processing (NLP) inference models that were submitted to the MLPerf v2.0 Datacenter Open division. These submissions demonstrated the power of Deci’s Automated Neural Architecture Construction (AutoNAC) technology, which automatically generated models dubbed DeciNets and DeciBERT, thus delivering breakthrough accuracy and throughput performance on Intel’s CPUs.
“We are excited to showcase another significant milestone in our journey to enable efficient deep learning inference on any hardware including resource constrained devices such as CPUs and other edge devices” said Yonatan Geifman, CEO and co-founder of Deci. “This major increase both in accuracy and throughput means that resource-intensive tasks that previously could not be carried out on CPUs are now possible, and in fact, will see a marked performance improvement. Hardware availability or compute power should never be a limiting factor for enterprises looking to employ the latest developments in deep learning.”
For their CV submission, Deci submitted three of its DeciNets models in the ResNet50 category under the offline scenario in the open division. Deci made submissions on two different hardware platforms: a 12-core Intel Cascade Lake CPU and two different Intel Ice Lake CPUs with 4 and 32-cores. Models were optimized on a batch size of 32 and quantized to INT8 using OpenVINO. Compared to the 8-bit ResNet50 model, Deci delivered +1.74% improvement in accuracy and 4x improvement in throughput.
Deci’s CV submission this year demonstrated a 37% improvement in throughput performance, as well as a notable improvement in accuracy, compared to their previous submission in 2020.
For NLP, Deci submitted its optimized DeciBERT models that produced outstanding results: accelerating question-answering tasks’ throughput performance on various Intel CPUs by 5x (depending on the hardware type and quantization level) while also improving the accuracy by +1.03%.
MLPerf gathers expert deep learning leaders to build fair and useful benchmarks for measuring training and inference performance of ML hardware, software, and services. The models submitted were optimized using Deci’s AutoNAC technology and quantized with Intel’s OpenVINO to 8-bit precision.
Deci’s AI-based AutoNAC technology automatically generates and optimizes deep learning architecture for any given data set and hardware to maximize its accuracy and inference performance. Deci’s AutoNAC technology, as well as its auto-generated DeciNets & DeciBERT, are ready for deployment and commercial use and can be easily integrated to support any CV or NLP task on a wide range of hardware types.
This announcement was originally published on Cision PRWeb.
Deci’s AutoNAC technology automatically generated a new family of models for image classification, dubbed DeciNets, for Intel’s Cascade Lake CPU, with these models outperforming well-known alternatives in both accuracy and runtime.
Tel Aviv, Israel, February 17, 2022 – Deci, the deep learning company harnessing Artificial Intelligence (AI) to build AI, today announced a new set of industry-leading image classification models, dubbed DeciNets, for Intel Cascade Lake CPUs. Deci’s proprietary Automated Neural Architecture Construction (AutoNAC) technology automatically generated the new image classification models that significantly improve all published models and deliver more than 2x improvement in runtime, coupled with improved accuracy, as compared to the most powerful models publicly available such as EfficientNets, developed by Google.
While GPUs have traditionally been the hardware of choice for running convolutional neural networks (CNNs), CPUs, already more commonly utilized for various computing tasks, would serve as a much cheaper alternative. Although it is possible to run deep learning inference on CPUs, generally they are significantly less powerful than GPUs. Consequently, deep learning models typically perform 3-10X slower on a CPU than on a GPU.
DeciNets closes the gap significantly between GPU and CPU performance for CNNs. With DeciNets, tasks that previously could not be carried out on a CPU because they were too resource intensive are now possible. Additionally, these tasks will see a marked performance improvement: by leveraging DeciNets, the gap between a model’s inference performance on a GPU versus a CPU is cut in half, without sacrificing the model’s accuracy.
“As deep learning practitioners, our goal is not only to find the most accurate models, but to uncover the most resource-efficient models which work seamlessly in production – this combination of effectiveness and accuracy constitutes the ‘holy grail’ of deep learning,” said Yonatan Geifman, co-founder and CEO of Deci. “AutoNAC creates the best computer vision models to date, and now, the new class of DeciNets can be applied and effectively run AI applications on CPUs.”
“There is a commercial, as well as academic desire, to tackle increasingly difficult AI challenges. The result is a rapid increase in the complexity and size of deep neural models that are capable of handling those challenges,” said Prof. Ran El-Yaniv, co-founder and Chief Scientist of Deci and Professor of Computer Science at the Technion – Israel Institute of Technology. “The hardware industry is in a race to develop dedicated AI chips that will provide sufficient compute to run such models; however, with model complexity increasing at a staggering pace, we are approaching the limit of what hardware can support using current chip technology. Deci’s AutoNAC creates powerful models automatically, giving users superior accuracy and inference speed even on low-cost devices, including traditional CPUs.”
In March 2021, Deci and Intel announced a broad strategic collaboration to optimize deep learning inference on Intel Architecture (IA) CPUs. Prior to this, Deci and Intel worked together at MLPerf, where on several popular Intel CPUs, Deci’s AutoNAC technology accelerated the inference speed of the well-known ResNet50 neural network, reducing the submitted models’ latency by a factor of up to 11.8x and increasing throughput by up to 11x.
Deci’s AutoNAC technology is already serving multiple customers across industries in production environments. To learn more about how DeciNets can benefit enterprises using CPUs to carry out inference tasks, visit here.
To learn more about how to boost deep learning models’ performance on CPUs, visit here.
This announcement was originally published on Cision PRWeb.
Deci is ISO 27001
Certified
from transformers import AutoFeatureExtractor, AutoModelForImageClassification
extractor = AutoFeatureExtractor.from_pretrained("microsoft/resnet-50")
model = AutoModelForImageClassification.from_pretrained("microsoft/resnet-50")