ParameterServer is an implementation for dmlc/ps-lite in java, which is a light and efficient implementation of the parameter server framework, and can be easily used to implementing distributed deep learning training framework, such as TensorFlow, PyTorch, MXNet, etc.
ParameterServer is built using Gradle. to build ParameterServer, run:
./gradlew build
This will run tests automatically, if you want build without running tests, run:
./gradlew build -x test
The jar required to run ParameterServer will be located in ./build/libs/
. run:
java -cp "build/libs/*" cn.daugraph.ps.Bootstrap
// TODO