Workflows

Workflows#

A workflow is a packaged way to run Isaac Lab somewhere other than a local Python environment. The task code, the configuration, and the command line stay the same; only the environment that executes them changes. Use this section once Isaac Lab runs locally through Installation and you need a reproducible environment, a job scheduler, or a remote GPU.

Every workflow below starts from the same container image, so they build on each other: the cluster workflow converts that image to Apptainer, and the cloud workflow provisions a machine that runs it.

Docker

Build and run Isaac Lab in a container for a reproducible environment on any Docker host. Start here – the other two workflows build on this image.

Running Isaac Lab in Docker
HPC clusters

Convert the container to an Apptainer image and submit jobs through SLURM, PBS, or OSMO.

Running on HPC clusters
Cloud workstations

Provision a GPU workstation on AWS, GCP, Azure, or Alibaba Cloud with Isaac Automator.

Cloud Deployment