MindSpore Tutorial

Quick Start

Basic processes of MindSpore deep learning, including data processing, model loading and saving, and graph mode acceleration.

Data Processing

Provide cases of data processing enhancement, cache, and pipeline functions.

Compilation

Provide MindSpore compilation syntax support cases and graph mode programming cases.

Parallel

Provide practice cases and optimization policies, such as data parallelism, operator-level parallelism, and optimizer parallelism.

Debugging and Tuning

Provide debugging cases for dump, profiler, and dryrun functions.

Custom Programming

Provide custom operator and customized convergence practice cases.

Inference

Describe the device-to-device process of MindSpore inference, including model building and weight segmentation.

High Availability

Describe the training HA feature, including fault recovery and graceful process exit.

Model Cases

Provide model migration guidance and various CV, NLP, and generative model building cases.