Skip to main content

Computational Resources

When creating a workflow each Piece can be considered a task that needs to be executed. As some tasks require more computational resources than others, Domino allows you to specify the resources needed for executing each one. For example, a Piece that trains a model can be executed on a GPU, while a Piece that performs a simple data cleaning can be executed on a CPU. Editing the computational resources of a Piece is done by double clicking on the Piece, going to Advanced Options and selecting the desired options.

The available options are:

  • CPU: The number of CPUs, in milicores(m), to allocate to the Piece.
  • Memory: The amount of memory, in mebibyte(Mi) to allocate to the Piece.
  • GPU: Define if should give access to a GPU to the Piece. To use a GPU, your cluster, if running in Kubernetes mode, or your machine, if running in compose mode, must have a GPU available. Also the piece must be built with a docker image that has the necessary drivers to use the GPU, you can find a base image with the necessary drivers here
info

CPU and Memory are considered by Domino only when running in Kubernetes.

Computational Resources