SuperGradients: Deep learning training library



Most of the time, enterprise AI developers struggle to develop and deploy production-ready deep learning models or deep learning training library. The overhead of integrating with various existing training tools and the effort to reproduce the training results for state-of-the-art models is time-consuming and causes headaches for beginners and experts alike. Various challenges affect the whole process, including data and model architecture selection.

However, Deci, a deep learning development company harnessing Artificial Intelligence (AI) to build AI, is releasing SuperGradients, an “all-in-one” deep learning training library for computer vision models.


“As deep learning is maturing and becoming more widely adopted, the model development and training processes must be simplified. To deliver on the promise of deep learning, teams require tools that can facilitate the architecture selection phase, as well as enable them to achieve better training results faster,” said Yonatan Geifman, CEO of Deci.

With SuperGradients, AI developers no longer need to spend time scouring through different repositories to find the most suitable architecture. SuperGradients contains all of the most commonly used models in one place. Additionally, the library has built-in access to pre-trained state-of-the-art models, many of which deliver higher accuracy compared to other libraries. Meaning developers can obtain better results faster, while easily integrating these models into their codebase.

SuperGradients also supports the easy fine-tuning and retraining of DeciNets. A family of state-of-the-art deep learning models generated by Deci’s proprietary Automated Neural Architecture Construction (AutoNAC) technology. DeciNets are currently being used by leading AI teams worldwide for multiple use cases.

 The library includes proven training recipes for easy reproduction of training results, thus making AI more accessible for everyone. In addition, users can easily load and fine-tune pre-trained state-of-the-art models (YOLOv5, DDRNet, EfficientNet, RegNet, ResNet, MobileNet, etc.). Of which many were optimized to deliver higher accuracy compared to existing training libraries, according to the vendor.

“By offering developers better tools to build, optimize and deploy models. We help to simplify the entire deep learning lifecycle. And enable developers to focus on what they do best- creating innovative AI solutions. To solve the world’s most complex problems,” said Yonatan Geifman.


Subscribe to our newsletter

Your emaill address should be use only for updating you on our articles, in the respect of the privacy law

Share post: