Good news! Amazon Cloud Technology DGL project won the 2021OSCAR open source peak case

"Follow" and "star" us,

Receive information about the Amazon cloud technology every day latest information !

On September 17, 2021, at the "2021 Open Source Industry Conference" hosted by China Academy of Information and Communications Technology,
Amazon Cloud Technology DGL won the "2021OSCAR Open Source Peak Case" -Open source community and open source project awards, demonstrating the technical strength and open source contribution of Amazon Cloud Technology.

At the annual open source industry conference, the China Academy of Information and Communications Technology will grandly announce the "OSCAR Open Source Pinnacle Case" award, which aims to establish a model of open source and better promote the implementation of open source technology in the Chinese market. In 2021, the award will be divided into five areas: "Open Source People", "Open Source Community and Open Source Projects", "Open Source Companies-Open Source Operations and Ecological Construction", "Open Source Companies-Open Source Governance", and "Open Source Innovation (Secondary Development)". . In the field of "open source communities and open source projects", selected more than 170 outstanding projects from many well-known companies through the cloud computing standards and open source promotion committee members. Amazon Cloud Technology DGL stood out in the fierce competition and succeeded Won national certification and recognition and won awards.

DGL (Deep Graph Library) is a dedicated deep learning framework for graph structure data. Open source was announced at the Neurips conference in December 2018. The development team originally came from NYU and NYU Shanghai, initiated by Professor Zhang Zheng, and is now mainly developed by the Amazon Cloud Technology Shanghai Artificial Intelligence Research Institute led by Professor Zhang Zheng. The project has received widespread attention and praise since its launch.

DGL is an open source project with far-reaching significance, it ease of use and strong , adopt a "map" as the core programming abstractions, natural near the semantic graph algorithms; high performance , fusion technology uses sub-count, Compared with similar systems, it can speed up 10 times or even dozens of times. In this way, enterprises can face complex business scenarios and improve the efficiency and accuracy of routine work such as speech analysis, text analysis, and time series analysis; multi-platform supports , and supports multiple deep learning frameworks (PyTorch, MXNet, TensorFlow), Users can perform large-scale graph neural network training based on this; massive giant graph supports , supports multi-machine multi-card distributed training, and can train billion-point graphs. DGL hopes to help developers put graph neural networks into specific business scenarios, improve business models, and create more value.

At the same time, DGL has a wide range of influence in the academic circle. There are nearly 400 academic papers published based on DGL. Many well-known scholars, including Turing Prize winner Yann Lecun, have spoken highly of the project. DGL has done special practical teaching in almost all top data mining conferences (KDD, WWW, WSDM, etc.). In just two years, the number of contributors to external projects from academia and industry has grown to more than 140. In addition, the DGL-based software ecology has also begun to take shape, deriving a variety of dedicated AI toolkits for molecular biology, knowledge graphs, natural language processing, recommendation systems, etc.

After DGL is open sourced, it also faces some challenges in terms of project maintenance and community building and takes active responses:

Challenge 1: Graph deep learning is a new and rapidly evolving field.

The DGL team has been actively collecting feedback from the open source community for effective backward iteration. And combined with the latest scientific research results, fast forward iteration.

Challenge 2: is widely used and has a strong domain due to the universality of graph structure data.

The team sets up stepped documents and tutorials on community operations to train developers in the field. Maintain communication with academia and industry to establish a win-win ecosystem.

(In terms of document construction, the original "document tower of Hanoi" management method)

is precisely these technical highlights of the project and the advantages of community operations that make DGL be rated as the open source community and open source project of the

future, Amazon Cloud Technology will continue to innovate, contribute more outstanding open source projects, and better support open source and integrate and integrate with cloud computing technology. It will accelerate open source-based machine learning, artificial intelligence and other technologies in many fields The application in production practice provides more mature growth opportunities for global users and enterprises.