Skip to content

Running Singularity containers from online repositories

Let us try to run a small and simple container from Docker Hub repository. Singularity, will pull the docker image in the cache, convert it and run it. The output should look something like this:

Important - reminder

Environmental variables that will help you to redirect potentially large folders to alternative location - keep in mind that your $HOME folder is relatively small in size.

# Workshop settings on Pelle
export APPTAINER_CACHEDIR=/proj/sinapp/nobackup/$USER/SINGULARITY_CACHEDIR
mkdir -p "$APPTAINER_CACHEDIR"


apptainer run docker://sylabsio/lolcow

INFO:    Converting OCI blobs to SIF format
INFO:    Starting build...
INFO:    Fetching OCI image...
27.2MiB / 27.2MiB [====================================================================================================] 100 % 8.5 MiB/s 0s
45.8MiB / 45.8MiB [====================================================================================================] 100 % 8.5 MiB/s 0s
INFO:    Extracting OCI image...
INFO:    Inserting Singularity configuration...
INFO:    Creating SIF file...
 ________________________________
< Thu Oct 2 09:35:31 Europe 2025 >
 --------------------------------
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||
The container executes predefined command date | cowsay | lolcat.

  • date - print current date and time.
  • cowsay - configurable speaking/thinking cow (and a bit more)
  • lolcat - rainbow coloring effect for text console display

Let's run it again. 

apptainer run docker://sylabsio/lolcow

INFO:    Using cached SIF image
 ________________________________
< Thu Oct 2 09:37:31 Europe 2025 >
 --------------------------------
        \   ^__^
         \  (oo)\_______
            (__)\       )\/\
                ||----w |
                ||     ||
Note, that singularity, after contacting the repositories, realizes that the container is in the local cache and proceeds to run it. But where is it?

More details...

apptainer cache list

There are 1 container file(s) using 87.96 MiB and 8 oci blob file(s) using 99.09 MiB of space
Total space used: 187.04 MiB

Info

Over time the cache will grow and might easily accumulate unnecessary "blobs". To clean the cache you can run. 

$ apptainer cache clean
Here is how the cache might look like: 
apptainer cache clean --dry-run
User requested a dry run. Not actually deleting any data!
INFO:    Removing blob cache entry: blobs
INFO:    Removing blob cache entry: index.json
INFO:    Removing blob cache entry: oci-layout
INFO:    No cached files to remove at /home/ubuntu/.apptainer/    cache/library
INFO:    Removing oci-tmp cache entry:     a692b57abc43035b197b10390ea2c12855d21649f2ea2cc28094d18b93360eeb
INFO:    No cached files to remove at /home/ubuntu/.apptainer/    cache/shub
INFO:    No cached files to remove at /home/ubuntu/.apptainer/    cache/oras
INFO:    No cached files to remove at /home/ubuntu/.apptainer/    cache/net

More examples

apptainer run docker://dctrud/wttr
output

Info

Coderefinery Excellent course material by the CodeRefinery project
Containers on HPC with Apptainer

Pulling Singularity container from online or local library/repository

  • library:// to build from the Container Library
    library://sylabs-jms/testing/lolcow
  • docker:// to build from Docker Hub
    docker://godlovedc/lolcow
  • shub:// to build from Singularity Hub
  • path to a existing container on your local machine
  • path to a directory to build from a sandbox
  • path to a Singularity definition file

Advanced

This is only an example how easy is to run python in a container with tensorflow installed on a computer with Nvidia GPU.

Tensorflow

Let's have some tensorflow running. First pull the image from docker hub (~2.6GB). 

apptainer pull docker://tensorflow/tensorflow:latest-gpu

INFO:    Converting OCI blobs to SIF format
INFO:    Starting build...
Getting image source signatures
...

If you have a GPU card, here is how easy you can get tensorflow running. Note the --nv option on the command line.

apptainer exec --nv tensorflow_latest-gpu.sif python3

Python 3.8.10 (default, Nov 26 2021, 20:14:08) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> tf.config.list_physical_devices('GPU')
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

NVIDIA Deep Learning Frameworks - additional material.