Docker Setup for EdgeAI TIDL Tools

January 23, 2026 ยท View on GitHub

This directory contains scripts and configuration files for setting up and running EdgeAI TIDL Tools in Docker containers. Docker provides an isolated environment with all the necessary dependencies pre-installed, making it easier to use the tools across different systems.

Overview

The Docker setup supports both CPU and GPU-based TIDL tools:

  • CPU Tools: Default configuration using Ubuntu 22.04 base image
  • GPU Tools: NVIDIA GPU-accelerated configuration using NVIDIA HPC SDK with CUDA support

Setup Instructions

The following environment variables can be exported to configure the docker:

Environment VariableDefaultAvailable OptionsNotes
TIDL_TOOLS_TYPECPUCPU or GPUSetting TIDL_TOOLS_TYPE=GPU sets up the docker image for tidl tools built with OpenACC based acceleration
REPO_LOCATIONUnsetANYRepo location for pulling docker image in case of a private network
PROXYUnsetANYPROXY to be used in case of a private network

Note: In case you are under TI's network, set REPO_LOCATION=artifactory.itg.ti.com/docker-public/library/ and PROXY=TI WEBPROXY

Docker image for CPU tools

1. Change CWD to repo root directory

cd <EDGEAI_TIDL_TOOLS_REPO_DIR>

2. Install Docker and Required Dependencies

TIDL_TOOLS_TYPE=CPU ./scripts/docker/setup_docker.sh

This script will:

  • Install Docker if not already installed
  • Add your user to the Docker group to run Docker without sudo

3. Build the Docker Image

TIDL_TOOLS_TYPE=CPU ./scripts/docker/build_docker.sh

The script will:

  • Build ubuntu22.04 docker image from Dockerfile with name edgeai_tidl_tools_x86_ubuntu_22

4. Run the Docker Container

TIDL_TOOLS_TYPE=CPU ./scripts/docker/run_docker.sh

The script will:

  • Mount the project root directory to /home/root in the container
  • Set up shared memory (4GB)
  • Start an interactive terminal session

Docker image for GPU tools

1. Change CWD to repo root directory

cd <EDGEAI_TIDL_TOOLS_REPO_DIR>

2. Install Docker and Required Dependencies

TIDL_TOOLS_TYPE=GPU ./scripts/docker/setup_docker.sh

This script will:

  • Install Docker if not already installed
  • Add your user to the Docker group to run Docker without sudo
  • Install NVIDIA Container Toolkit

2. Build the Docker Image

TIDL_TOOLS_TYPE=GPU ./scripts/docker/build_docker.sh

The script will:

  • Build nvhpc based ubuntu22.04 docker image from Dockerfile_GPU with name edgeai_tidl_tools_x86_ubuntu_22_gpu

3. Run the Docker Container

TIDL_TOOLS_TYPE=GPU ./scripts/docker/run_docker.sh

The script will:

  • Mount the project root directory to /home/root in the container
  • Set up shared memory (4GB)
  • Start an interactive terminal session
  • Enable GPU access if using GPU tools
  • In case Docker is not able to detect GPU during run_docker, please install nvidia-container-toolkit

Notes

  • The container mounts the project root directory, so any changes made inside the container to files in this directory will persist outside the container.
  • For GPU support, ensure you have the NVIDIA Container Toolkit installed and configured properly.
  • The Docker images include all dependencies required for EdgeAI TIDL Tools development and usage.