Quick Start

Overview
Quick Start
Tensor
Data Loading and Processing
Building a Network
Automatic Differentiation
Model Training
Saving and Loading the Model
Accelerating with Static Graphs

Data Processing

Data Processing Overview
Data Loading and Sampling
Data Operation/Data transformation
MindRecord Format Conversion
Optimizing the Data Processing

Compilation

Introduction to Graph Mode Programming
Graph Mode Syntax - Operators
Graph Mode Syntax - Python Statements
Graph Mode Syntax - Python Built-in Functions
Graph Mode - Programming Techniques
Custom Fusion Strategy

Parallel

Distributed Parallelism Overview
Distributed Parallel Startup Methods
Data Parallel
Operator-level Parallelism
Optimizer Parallel
Pipeline Parallel
Optimization Techniques
Distributed High-Level Configuration Case
- Multi-dimensional Hybrid Parallel Case Based on Double Recursive Search

Debugging and Tuning

Dynamic Graph Debugging
Using Dump in the Graph Mode
Feature Value Detection
Ascend Performance Tuning
Error Reporting Analysis
DryRun

Custom programming

Custom Operators
Custom Backend
Custom Pass
Hook Programming

Infer

MindSpore Inference Overview
MindSpore LLM Inference with Framework
MindSpore Lite Inference Overview

High Availability

Fault Recovery
Training Process Exit Gracefully

Orange Pi

OrangePi AIpro Development
Environment Setup Guide
Model Online Inference
Quick Start

Model Cases

Model Migration
Computer Vision
Natural Language Processing
Generative

Tutorials

»
Distributed High-Level Configuration Case
View page source

Distributed High-Level Configuration Case

Multi-dimensional Hybrid Parallel Case Based on Double Recursive Search

Previous Next

© Copyright MindSpore.

Built with Sphinx using a theme provided by Read the Docs.