hyperai/tvm-cn

Apache TVM is an open-source deep learning compiler. Its purpose is to help machine learning engineers take a trained model and run it efficiently on a wide range of hardware, including standard CPUs, NVIDIA and AMD GPUs, ARM processors, and other embedded chips. Rather than writing separate optimization code for each hardware target, developers can use TVM to compile and tune a model once and then deploy it broadly. This repository is a Chinese-language translation project for the official TVM documentation. The translators noted that Chinese-language TVM learning resources were scattered and difficult to learn from in a structured way, so they created this centralized translation to make TVM more accessible to Chinese-speaking developers and researchers. The translated documentation is based on TVM version 0.10.0. The team updates the Chinese docs as the official TVM documentation evolves. Anyone who finds a translation error or an ambiguous phrase can file an issue or submit a pull request on GitHub. The documentation site is published at tvm.hyper.ai and readers can browse it without any local setup. For contributors who want to run the site locally, the project uses Docusaurus, a documentation framework that builds static HTML sites from Markdown files. Setup requires Node.js and the pnpm package manager, and the standard development workflow runs a local preview server. The README also explains how to handle images: external image links should be downloaded and stored inside the repository rather than left as remote URLs, to avoid broken images if the original source moves. The repository is primarily a community resource rather than a software library. Its value is the translated documentation itself and the ongoing effort to keep it current with upstream TVM releases. Versioning is managed through Docusaurus's versioning system, so older TVM documentation versions remain accessible alongside the latest.