check port forwarding first

1. run server :: preset epoch, client_k which you want aggregate some
2. run client :: preset epoch same as server, gpu num which you want to put in (if can't, switched to cpu )

• Server

• DEV
• new docker container init
• will be added with new dockerfile
• dockerfile-build

docker build -t hail/pinme/server:1.5 server/

• dump sectet

docker commit --change "ENV SERVER_POT=0000"

• dockerfile -run

docker run -d --name pinme.server hail/pinme/server:1.5

• shell

source server/start.sh

• python directly

python run_server.py  --SERVER_PORT 59919 --SERVER_HOST '127.0.0.2' --model fedavg --k_clients 1

• Client

• new dockerfile

docker build -t hail/pinme/client:1.5 client/

• dockerfile -run

docker run -d --name pinme.client1 hail/pinme/client:1.5 

• shell

source clinet/start.sh

• python cmd

python run_client.py --CLIENT_PORT 59919 --CLIENT_HOST '127.0.0.2' --model fedavg --n_epochs 1 --gpu 0

• if you want to use server & edge, not locally, then set server's port & host to fit in docker forwarding setting

• @base docker

• @guide

• @cifar-10

• communicate : communicate stream & copy stream
• models : model architecture & federate learning
• worker : data & model & network setter, data loader
• fed : frame-work hugging phase
• server : server class & run script
• clinet : client class & run script
• utils : debug, plot, logger

@software{pinME,
  author = {Hail},
  month = {12},
  title = {{Frame-work for federate learning}},
  url = {https://github.com/Hail-cali/pinME},
  version = {1.2.0},
  year = {2021}
}