As a federated learning framework, FATE supports many algorithms. It is also the needs of many enterprises to select the corresponding algorithm and accelerator card according to the business. Currently, KubeFATE supports the selection of algorithms and accelerator cards.
Whether in docker-compose or k8s deployment, the following two parameters can be selected:
algorithm
Algorithm choicedevice
Computing Device Selection
The choice of algorithm consists of two options:
Basic
Basic is the default option, which includes dependencies related to the removal of nn (including homo_nn and hetero_nn) algorithms.NN
NN contains all the dependencies required for nn to include (homo_nn and hetero_nn). NN can only be used when computing is Eggroll
Device selection consists of an option:
CPU
The CPU is a computing device that uses the CPU as a FATE computing device.