Standalone_Deployment.md

Standalone Deployment Guide

Standalone mode deployment is divided into EFLS-data and EFLS-train modules.

Environment requirements:

• docker

EFLS-data

docker image build

cd ${EFLS}/efls-data
sudo docker build -t efls-data:v1 -f ./Dockerfile ./

test

After building the image, you can run a test file inside the container generated by the image to test the feasibility of local data join. The local test will randomly generate two copies of tfrecord data, and then run the server and client. The client uses grpc communication to transmit data to the server, and the server returns the result of intersection.

docker run -it efls-data:v1 bash 
python /xfl/test/test_data_join.py

Expected output "OK"

quick start

Users can use their own data or modify /efls-data/test/data_maker.py to generate data for sample intersection.

We provide scripts for local data intersection, which is located in the /xfl/bin directory in the container. Please refer to documentation to get more information.

EFLS-train

docker image build

cd ${EFLS}
sudo docker build -f docker/Dockerfile.efls-train -t efls-train:v1 .

test

After building the image, you can run the test file in the container generated by the image to test the feasibility of local training. The test file is in /tmp/efl/python/efl/example.

docker run -it efls-train:v1 bash
cd /tmp/efl/python/efl/example/mnist
python generate_data.py
python leader.py --federal_role=leader &
python follower.py --federal_role=follower

quick start

Users can modify according to the example to design a new training model, or refer to the corresponding document.