dockgpu-template

Template repository to create a container with GPU support and CUDA, cuDNN, TensorRT libraries. The container is based on the nvidia/cuda image.

Note: This template is intended to be used with vscode-dev-containers extension.

Included Features

ZSH is installed and made as the default shell. Some useful plugins are installed such as:

- Syntax-Highlighting
- Autosuggestions
- Interactive-cd
- Zoxide

Other installed tools are:

- Git, gh
- gcc, g++, make
- jdk, jre, maven
- openssl, curl, wget
- mamba, conda

Usage

To use this template edit theese files as described:

• .env: Set the container, image and user name.
• devcontainer.json: Substitute dockgpu as you want and dock username with the one you set in .env file. Theese are needed to set correctly the vscode remote container extension.
• Dockerfile: Substitute dock username with the one you set in .env file.

Notes

Inside the dockerfile you can find a RUN command to launch a script that downloads a file froma a OneDrive link. This is a workaraund to get the cudnn.h file that seems to be not available on the official nvidia/cuda image.

Installed python packages are listed in the last RUN command of the dockerfile. You can add or remove packages as you need.