Skip to content

Latest commit

 

History

History
24 lines (15 loc) · 994 Bytes

FATE_Algorithm_and_Computational_Acceleration_Selection.md

File metadata and controls

24 lines (15 loc) · 994 Bytes

FATE Algorithm and Computational Acceleration Selection

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 choice
  • device Computing Device Selection

Algorithm

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

Device selection consists of an option:

  • CPU The CPU is a computing device that uses the CPU as a FATE computing device.