Version 2.0 of the Deci Deep Learning Platform makes it easier than ever for AI developers to build, optimize, and deploy computer vision models for any hardware and environment, including the cloud, peripherals, and mobile devices with exceptional runtime accuracy and performance.

AI developers are facing a difficult struggle in developing ready-to-produce deep learning deployment models. These challenges can be to a large extent is due to the gap in AI efficiency faced by the industry, where algorithms are becoming more powerful and complex, but the available computing power is not keeping pace. This gap also creates financial barriers, making the development and processing of in-depth training more cumbersome and expensive.

Although the demand for neural architecture (NAS) is presented as a potential solution for automating the design of superior artificial neural networks that can surpass manually designed architectures, the resource requirements for working with such technology are excessive. To date, NAS has been used successfully only by technology giants such as Google, Microsoft and academia, which proves its impracticality for most developers.

To solve this problem, Deci’s platform, powered by its own NAS engine called AutoNAC (Automated Neural Architecture Building), allows AI developers to automatically and affordably build efficient computer vision models that provide the highest accuracy for any given hardware, speed, size and purpose. The models generated by Deci are superior to other known state-of-the-art (SOTA) architectures by a factor of 3x-10x.

Developers can start their projects with pre-trained and optimized models (DeciNets) 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 tailored to their needs. specific use – cases. In addition, the platform supports teams with a wide range of tools needed to develop deep learning applications, including a hardware-aware zoo model for easy selection and benchmarking of models and hardware, SuperGradients – an open source training library with proven recipes for faster learning, automated optimizations during execution, model packaging, etc.

Using the Deci platform, AI developers achieve improved output performance and efficiency to enable deployment of low-end endpoints, maximize hardware utilization, and reduce training and output costs. The whole development cycle is shortened and the uncertainty about how the model will be deployed on the output hardware is eliminated.

With Deci’s hardware-oriented zoo model, developers can quickly measure the lead time of pre-trained and optimized models of and various hardware, including end devices, through Deci’s SaaS platform. Simplify the process of choosing hardware and model by eliminating the need to manually adjust and test different combinations of models and hardware.

With the Deci version 2.0 platform, AI developers can:

Automatically find accurate and efficient architectures tailored to your application, hardware and performance with Deci’s AutoNAC engine.

Take advantage of proven recipes for hyperparameters with the PyTorch-based open source Deci learning library called SuperGradients.

Automatically compile and quantize your models and evaluate various production settings.

Developers can deploy their deep learning workloads in any environment with Decy’s python-based inference machine.

The Deci platform includes three levels:

Free community level: For data scientists and machine learning engineers who want to find the best models, simplify hardware evaluation and increase runtime productivity.

Professional level: For in-depth training teams that want to quickly achieve productivity of the production class output and shorten development time.

Enterprise Tier: For deep learning experts who want to achieve specific performance goals for highly customized uses.

Deci, Hugi Tower 8F, Abba Hillel Silver Rd 12, 5250606, Ramat Gan, Israel,,

DesignFast Banner version: 68c59059

Deep learning development platform speeds AI work

Previous article100V 1.7mΩ GaN power transistor in 10.2mm2 fingerprint
Next articleXiaomi Mi 12 Ultra with Surge C2 ISP should come in July