OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Infrastructure & Serving

Tnn

Tnn is an open-source C++ neural-network inference framework for deploying trained models in mobile, desktop and server applications. Its workflow includes converting models to TNN format and compiling an inference engine for a target platform.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Tnn — OpenViST project profile in AI Infrastructure & Serving.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Tnn when you have a trained model and need a framework workflow for preparing and integrating inference into an application for a mobile, desktop or server target.

Who it is for

  • Developers integrating neural-network inference into mobile applications.
  • Teams building desktop or server software that runs trained models.
  • Model developers seeking a conversion and target-specific engine-build workflow.

Documented uses

  • Convert a trained TensorFlow, PyTorch or Caffe model into TNN format.
  • Compile an inference engine for a target platform and selected hardware acceleration option.
  • Call Tnn from application code to run inference.
  • Develop applications for tasks such as face detection, OCR, object detection or pose estimation using the framework.
Before you choose

Requirements & limits

The repository identifies Tnn as open source.

  • The documented workflow requires model conversion before engine compilation and application inference.
  • The engine build is target-specific; listed acceleration choices include ARM, OpenCL, Metal, NPU, X86 and CUDA.
  • The README emphasizes mobile support and also describes desktop and server use.
  • The repository provides examples across several inference tasks, including face detection, OCR and object detection.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare a trained model for Tnn inference

Follow the documented conversion and engine-compilation stages to prepare a TNN model for a target platform.

Before you start

  • A trained model from TensorFlow, PyTorch or Caffe.
  • A target platform and hardware acceleration option supported by the project documentation.

Steps

  1. Convert the trained model into TNN format using the project’s model-conversion tools.

  2. Compile the TNN engine for the target platform, selecting an acceleration option such as ARM, OpenCL, Metal, NPU, X86 or CUDA according to hardware support.

Expected outcome

A converted TNN model and a compiled inference engine for the chosen target, ready for the subsequent application-inference stage.

Guide limitations

  • This guide covers model conversion and engine compilation only; it does not document application integration or a successful inference run.
  • The acceleration options depend on hardware support.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
Tencent/TNN · Official homepage title · retained in its source language
How we review sourcesSuggest a correction