Experiment configuration#

The general experiment configuration are related to how the experiment is run. Specifically, users can define:

  • mode: the way of running the experiment. It can be in a federated or a centralized fashion. Note that only the federation mode is implemented for the moment;

  • train: for performing the training step or not;

  • seed: the seed for the random number generator, which is essential for the experiment reproducibility. This does not apply to the data pipeline as it exists a second seed for this step;

  • device: the device where the training and evaluation will be performed. The supported settings are cpu, cuda,mps, or auto;

  • inmemory: whether to use caching to save memory. If true, the data is stored in memory, otherwise it is stored on disk.

See also

For more information about device, seed, and inmemory, see Experiment configuration from Fluke documentation.

This Page