Documentation

MindStudio

MindStudio is a full-process toolchain provided by Huawei for Ascend AI developers. It is dedicated to providing an end-to-end Ascend AI application development solution, enabling developers to efficiently complete training development, inference development, and operator development.

  • What Is MindStudio

    Allows you to quickly understand the functional architecture, development scenarios, and presentation modes of MindStudio.

  • Release Notes

    Describes the CANN version that matches MindStudio, changes in the current version, and resolved issues.

Getting Started

  • Developer Journey

    This section describes the overall process of the development tool in the operator, training, and inference scenarios.

  • Operator Development Tool Quick Start

    The following uses the single-operator API calling mode as an example to describe the entire process of using the operator development tool.

  • PyTorch Training Tools Quick Start

    This section uses the PyTorch training scenario as an example to describe how to use the migration, accuracy, and performance tools.

  • Inference Tool Quick Start

    Describes the applications of tools in the foundation model inference toolchain, such as model compression, inference data dump, automatic accuracy comparison, and performance tuning.

Operator Tools

  • Operator Development Tools

    Describes how to use the operator development tool sets (such as msKPP, msOpGen, msOpST, msSanitizer, msDebug, and msProf).

  • Operator compilation tool

    Provides a command line tool for compiling operators to generate operator binary files.

Training Inference Development Tool

  • ATC tool

    Converts a network model into an offline model in .om format supported by the Ascend AI Processor.

  • AOE tool

    The automatic optimization tool aims to make full use of hardware resources and improve network performance.

  • Analysis and Migration Tool

    Migrate the PyTorch training script to the Ascend NPU in one-click mode.

  • Accuracy Debugging Tool

    Provides accuracy comparison in TensorFlow training scenarios to help quickly locate model accuracy problems.

  • Profiling Tools

    Profile data collection and analysis in scenarios such as PyTorch and TensorFlow training.

  • HCCL Performance Tester Tool

    Test the correctness and performance of the HCCL collective communication.

  • AMCT Tool

    Model compression tool package, which provides multiple model compression features such as quantization and tensor decomposition.

  • Operator and Model Query Tool

    Query the models and operators supported by CANN of the current version.

GUI-based development tool

  • Installation Guide

    Describes how to download and install MindStudio IDE.

  • IDE User Guide

    MindStudio provides functions such as project management, model development, application development, operator development, and debugging and optimization.

Cases

References