There is a set of container images that you can use in your computational workflows without building your own. The images are based on Docker Stack project with NVIDIA CUDA libraries added, so that you could use those with GPUs. You can get the list of software included in the images on the Docker Stack web page.
The list of images and registry links are available in our GitLab registry.
The PRP image (gitlab-registry.nrp-nautilus.io/prp/jupyter-stack/prp) is based on Tensorflow one with addition of:
astropy bowtie fastai keras nbgitpuller opencv-python psycopg2-binary tensorflow-probability torch torchvision visualdl git+https://github.com/veeresht/CommPy.git
We can add more libraries to this image on request in Matrix.
The Desktop image has X Window system installed and you can launch the GUI interface in Jupyter with this image. It’s based on Minimal stack one.
If you’re using these images in your jobs, you can speed up the image download time by using Kraken.