Project Introduction

Apache TVM is an open-source machine learning compiler framework for CPUs, GPUs, and machine learning accelerators, designed to enable machine learning engineers to efficiently optimize and run computations on any hardware backend.

This project is a complete TVM learning tutorial, covering all aspects from beginner to advanced, including model import, compilation optimization, operator tuning, and microcontroller deployment.

The TVM version in this tutorial was updated in January 2026. tvm==0.22.0

Table of contents

I. Introduction to Basics and Core Modules

• 00_quick_start_cn.ipynb Quick Start: This tutorial is for Apache TVM beginners.
• 01_ir_module_cn.ipynb IR module: The core abstraction layer of Apache TVM Unity

II. TVM Usage and Optimization Practices

• 00_e2e_opt_model_cn.ipynb End-to-end model optimization
• 01_customize_opt_cn.ipynb Custom optimization
• 02_optimize_llm_cn.ipynb LLM optimization (This chapter requires working with remote devices and cannot be run directly on the platform)
• 03_cross_compilation_and_rpc_cn.ipynb Cross-compilation and RPC
• 04_export_and_load_executable_cn.ipynb Exporting and loading the Relax executable

license

This tutorial is licensed under the Apache License 2.0 for the Apache TVM project.