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

Autokernel

Autokernel is a C++ toolset that uses Halide to generate low-level neural-network operator implementations, search for optimized schedules on CPU and GPU, and integrate generated operators into Tengine.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Autokernel — 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 Autokernel when you need to generate or optimize low-level neural-network operators for a supported hardware target, or when your deployment workflow uses Tengine’s documented plugin integration.

Who it is for

  • Developers optimizing neural-network operators for deployment on different hardware.
  • Teams using Halide scheduling or integrating generated operators into Tengine.

Documented uses

  • Generate optimized low-level code from an operator algorithm description.
  • Search for Halide schedules targeting CPU or GPU.
  • Integrate generated operator code into Tengine without changing its core codebase.
  • Use a containerized CPU, CUDA or OpenCL environment with Halide and Tengine installed.
Before you choose

Requirements & limits

The repository is public and licensed under Apache-2.0. It provides CPU, CUDA and OpenCL Docker images; CUDA instructions specify nvidia-docker.

  • AutoSearch is described as still under development.
  • The documented integration path is for Tengine; the evidence does not confirm equivalent support for other deployment runtimes.
  • The repository provides Docker images for CPU, CUDA and OpenCL, and specifies nvidia-docker for CUDA.

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

Start the AutoKernel CUDA Docker environment

Open a shell in the documented AutoKernel CUDA image.

Before you start

  • Access to the AutoKernel Docker image.
  • Use of nvidia-docker for the CUDA image, as specified by the repository.

Steps

  1. Pull the CUDA image with nvidia-docker pull openailab/autokernel:cuda.

  2. Start an interactive shell with nvidia-docker run -it openailab/autokernel:cuda /bin/bash.

Expected outcome

An interactive Bash shell opens in the AutoKernel CUDA container.

Guide limitations

  • The repository specifies nvidia-docker rather than docker for the CUDA image.
  • Opening the container does not document how to generate or deploy an operator.
Documentation 1 ↗
Evidence and transparency

Sources & review

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