Mainstream framework support

The BML platform provides machine learning and in-depth learning development capability, supports mainstream development frameworks and supports the installation of third-party software packages to custom other frameworks. PaddlePaddle is an in-depth learning framework self-developed by Baidu AI Cloud platform. The BML platform always maintains support for its latest version.

Multiple modeling methods

BML platform provides multiple modeling methods, such as, interactive (Notebook), visualization (drag and drop components), task-based (task modeling), production line (intelligent visual ), which is suitable for users with different R & D capabilities to quickly implement model training, evaluation and prediction.

Full-cycle model management

The BML platform provides full-cycle service for model production and management, including data storage, data annotation, data processing, model training, model prediction, service monitoring, management of operation and maintenance of refined scheduling and user management. It provides model training platforms and mature vertical model production lines to support quick development and supports introduction and management of third-party models.

Open product feature interface

The BML platform provides Open API/SDK interface to facilitate the seamless connection of upper-level business applications. Product design has clear levels and interfaces are open, which is conductive to effective connection with existing IT work environments such as customers’ private cloud environments, local servers, big data platforms and operation and maintenance platforms.

Domestic Kunlun chip

Kunlun chip adopts the advanced AI architecture of Baidu, which is well-suited for cloud computing requirements of commonly used deep learning and machine learning algorithms, adaptive to the computing needs of various terminal scenarios. BML launched the Kunlun chip package in Suzhou area, seamlessly supporting Baidu PaddlePaddle framework by providing excellent performance and efficiency.