Product Introduction
With the rapid development of the AI industry, more and more enterprises have released deep learning platforms. These platforms provide comprehensive functions for model development, including dataset management, model training, model management, model deployment, and model inference, accelerating the delivery of models required by AI services.
MindX DL is a set of deep learning components that support data center training and inference hardware powered by Ascend AI Processors (in this document, NPU = Ascend AI Processor). It offers basic functions such as NPU scheduling, cluster performance test, and model protection, providing underlying software support for upper-layer model training, evaluation, deployment, and inference. It reduces the workload for developing underlying resource scheduling software for deep learning platform vendors and quickly enables partners to develop MindX DL-based deep learning platforms.
As shown in Figure 1, the MindX DL components are used to enable third-party deep learning platforms to complete training and inference. The following describes the component features.
- Cluster scheduling: enhances NPU scheduling based on Kubernetes and checks the NPU and node status.
- Model protection: provides encryption and decryption functions throughout the model lifecycle.
- Toolbox: provides functions such as bandwidth test, computing power test, and power consumption test, for standard PCIe cards, board cards, and modules of Atlas products.