Installing Parabricks via the RedHat/RPM bare metal installer requires:
A working CUDA installation
An active Anaconda environment
Access to the NVIDIA NGC Licensing Portal
You can test whether you have CUDA installed by running the following:
nvidia-smi
If the command is not found, you do not receive a valid status message for your GPU or your CUDA version is older than:
CUDA 10.1 for non-Ampere GPUs
CUDA 11.0 for Ampere GPUs
then you will need to install CUDA. Instructions for doing so can be found at https://developer.nvidia.com/cuda-downloads. Once you have installed CUDA, you may proceed with the following steps.
To request a trial license, go to https://www.nvidia.com/en-us/clara/genomics/ and click
on the "FREE 90-DAY TRIAL" button. You should receive two emails with additional instructions.
One tells you how to log on to NGC, the other gives you a direct link
to NGC (you'll need to be able to log in to NGC). Following the direct link will let you download
"Bare Metal RPM Package" resource, which will result in you downloading a file called
files.zip. Once you have downloaded files.zip, go through the following steps:
# Unzip the downloaded file. This extracts parabricks.rpm from files.zip
$ unzip files.zip
# Install the software.
$ sudo dnf install -y ./parabricks.rpm
# Verify your installation.
# This should display the parabricks version number:
$ pbrun version
During the installation you'll be asked
To accept the end user license agreement.
If you want to create a symbolic link from the Parabricks installation to
/usr/bin/pbrun(requires :code`sudo` access).If you want to install the Ampere version or the non-Ampere version.
To check if your system has Ampere GPUs, run the following:
nvidia-smi --query-gpu=name --format=csv,noheader
Any GPU labeled with the first letter A (e.g. 'Tesla A100', 'Tesla A40') is an Ampere GPU.
If your machine has more GPUs than you have licenses, you will need to specify which, and how many, GPUs to use. For example, use the following flags to select only the first two GPUs:
CUDA_VISIBLE_DEVICES="0,1" pbrun [command and inputs] --num-gpus 2
After installation, the pbrun command will start any tool in the Parabricks software suite.
Answering "yes" to the second question (about creating a symbolic link) will make pbrun
available for system-wide access. Otherwise, you can access pbrun from your local
installation directory, which is /opt/parabricks/pbrun by default.
To obtain to a permanent node-locked license, run the following command on your GPU server:
$ nvidia-smi -L
Send the output to parabricks-support@nvidia.com. We will send you a permanent license.bin file,
locked to the GPU(s) on that server.
Copy the new license.bin file to /opt/parabricks/license.bin to use your new license.
The license.bin file must be readable by the user running the software.**