Skip to content

Software Packages

Conventional Operations

For using conventional tools such as conda, VSCode, pytorch, tensorflow, or containers with the Galvani cluster, please read through the Tutorials in this User Guide for instructions.

Development Tools

To see the development tools installed on Galvani run the following:

scl list-collections
It should output in terminal the following packages:

gcc-toolset-10
gcc-toolset-11
gcc-toolset-9
To source and enable one of these, run:

source scl_source enable <devtoolset-x>

Selecting a CUDA version

You can find a list of the different cuda versions on the compute nodes under:

ls -1 /usr/local/cuda* -d

which should provide terminal output like this:

/usr/local/cuda-11
/usr/local/cuda-11.7
/usr/local/cuda-11.8
/usr/local/cuda-12
/usr/local/cuda-12.1

You can select a specific CUDA version to use (e.g. cuda 11.8) by running (replace 11.8 with the desired CUDA version): 

export CUDA_VER=11.8 ; export PATH=/usr/local/cuda-$CUDA_VER/bin:$PATH ; export LD_LIBRARY_PATH=/usr/local/cuda-$CUDA_VER/lib64:/usr/local/cuda-$CUDA_VER/extras/CUPTI/lib64:$LD_LIBRARY_PATH
which will place the appropriate version in your PATH and LD_LIBRARY_PATH for this session. You can verify this worked by running nvcc --version and reading the output.

If you want to select the same CUDA version every time, please add the above command to your .bashrc file or sbatch file.

Last update: September 9, 2024
Created: September 9, 2024