Skip to content

Getting applications

To run applications on AI-LAB, you must use container images. On AI-LAB we use the container software, Singularity.

What is a container image?

A container image is a static, portable file that contains all the components needed to run a piece of software, including the code, runtime, system tools, libraries, and settings.

Pre-downloaded container images

The most straightforward method to acquire container images on AI-LAB is by accessing pre-downloaded container images stored in the /ceph/container directory. We aim to consistently update these container images to the latest versions.

You can check which container images exist in the /ceph/container directory on AI-LAB with ls:

ls /ceph/container

To use the container images, you can use them straight from the /ceph/container directory by referencing the absolute path, e.g.:

/ceph/container/pytorch/pytorch_24.09.sif

Download container images

Alternatively, you can access a wide array of container images by visiting NVIDIA GPU Cloud (NGC) and exploring whether NVIDIA provides a container image for the application you require. Refer to our guide here for detailed instructions.

Build your own container images

You also have the flexibility to create your own container images tailored to your specific environment requirements. Refer to our guide on building your own container image.


Now that you know how to obtain applications, let's delve into running jobs