Requirements to test and wrap gpu tool for Galaxy

Installation

1. ensure nvidia-smi is installed
2. ensure nvcc is installed
3. ensure the nvidia driver is installed on windows if ubuntu WSL is used

Install Graphics Card Drivers

Visit the official NVIDIA website to download NVIDIA DRIVER and fill in the fields with the corresponding graphics card and OS information. Download the driver and run the file to install the driver on Windows OS.

Install NVIDIA drivers in WSL

Start the Ubuntu terminal and run the folowing command to update the packages.

sudo apt update && upgrade

Then, install the nvidia drivers according to the version. In this example, we installed version 555 on ubuntu 22.04.

wget https://developer.download.nvidia.com/compute/cuda/12.5.0/local_installers/cuda_12.5.0_555.42.02_linux.run
sudo ./cuda_12.5.0_555.42.02_linux.run
or
sudo apt install cuda-toolkit-12-5

Verify the installation was successful, run the following command.

nvidia-smi

Note: Nvidia driver version 570 is backward compatibility for version 555. Verify NVCC - the CUDA compiler-driver tool that is installed with the CUDA toolkit, will always report the CUDA runtime version that it was built to recognize. It doesn't know anything about what driver version is installed, or even if a GPU driver is installed.

nvcc --version

Install planemo

Follow the instruction in planemo to install planemo

wrap a gpu tool and run planemo test. In this example, the dorado tool is a docker container

planemo test --docker --docker_run_extra_arguments ' --gpus all ' dorado-test-basecaller.xml

CUDA device requested but no devices found.

Ensure both PATH and LB_LIBRARY_PATH are set in bashrc

export PATH=$PATH:/usr/local/go/bin:/usr/local/cuda-12.5//bin
export LD_LIBRARY_PATH=/usr/local/cuda-12.5/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}