Docker Server Setup

The Machine Learning version of XNAT relies on the Docker system for both the short-term transfer learning and inference tasks as well as persistent containers for XNAT itself and other required components such as the PostgreSQL database. Production work will require a computing environment with sufficient memory, CPU and GPU resources. A specific example is the amount of memory reserved for the Docker server. You can run the software stack on a desktop system, but the default memory allocation for the Docker server will be too low to support all of the components. In our development environments, we use:

ResourceDocker DesktopDocker Server
Memory16 GB64GB
Swap1 GB6GB