beat.web 1.4 allows the use of GPU for algorithm. This milestone lists what is needed to do in order to use that technology on the platform.

On local machine:

• Create a new docker environment with GPU enabled packages like pytorch
• Validate starting of GPU enabled conda containers with nvidia-docker2
• Write sample algorithms that makes use of GPU
• Validate running an experience from beat.cmdline
• Validate starting and running an experience from beat.web

In infrastructure:

• Setup compute node to use GPU
• Test Docker with GPU
• Document setup procedure
• Validate functionality on compute node
• Test on infrastructure
• Deploy in production