Deep Learning Enterprise Architecture
Use immediately

Deep Learning Enterprise Architecture

MindSpore Deep Learning Framework

Mindspore FrontEnd Expression

Python API

Training/Inference/Expoert

Data Processing

Training/Inference/Expoert

MindSpore IR

GHLO

High Level Optimization

Auto Parallel

Auto Differentiation

Mindspore Graph Engine (Ascend/GPU/CPU Support)

Grpah Execution

Training/Inference/Expoert

Distributed Libs (Comms/PS)

Mindspore Backend Runtime (Cloud/Edge/Mobile)

CPU

GPU

Ascend 310

Ascend 910

Android/iOS

GLLO

Low Level Optimization

Pipeline Parallel

27
0
1
publish time: 2021-07-16
Kiraaaa

MindSpore is an open-source deep-learning training/inference framework used for mobile, edge, and cloud scenarios. This framework is designed to provide development experience with a friendly design and efficient execution for the data scientists and algorithmic engineers, native support for Ascend AI processors, and software-hardware co-optimization. It should be noted here that this open-source deep learning inference is also utilizing the cloud-native ecosystem for deployment and management. The below architecture diagram is created in EdrawMax and shows how MindSpore is looking forward to enabling users to use Jupyter to develop models. In the coming years, users can use Kubeflow tools like fairing to build containers and create Kubernetes resources to train their MindSpore models.

See More Related Templates