Skip to content

Latest commit

 

History

History
60 lines (42 loc) · 1.96 KB

README.md

File metadata and controls

60 lines (42 loc) · 1.96 KB
Raw

BZKF-Demo

This is a minimal MNIST Tensorflow Demo for the BKFZ Summer School with the PHT-meDIC.

Analyst should upload the algorithm analysis-train.py as entrypoint in the central UI. Admins should upload the corresponding data to their station.

If you are in the second round in thegr group Admins, you dont need to upload the data to the station again!

Analysts

The analyst instructions you can find in this link

Analyses code

This code loads a prepared dataset and partial trains a MNIST demo. Furthermore it uses Paillier encryption to securely count the number of training and testing samples (globally and privacy preserving)

Cheat sheet

  1. Download the App
  2. Create and upload both keys
  3. Submit an analysis
  4. Download and decrypt results

Please ask for help if you have in any of those steps issues.

Admins

Your instructions you can find in this link

Cheat sheet

  1. Create keys
  2. Login to the VM
  3. Install a PHT-meDIC station
  4. Approve and run analysis

Please ask for help if you have in any of those steps issues.

To copy demo train data to a station (replace YOUR_KEY, YOUR_DATA_PATH, STATION_NUM, IP_VM) and run this command:

scp -i ~/.ssh/YOUR_KEY -r /YOUR_DATA_PATH/ds_STATION_NUM ubuntu@IP_VM:/home/ubuntu/station/station_data/

To run the train use this command (replace STATION_NUM and TRAIN_ID):

{"repository":
    "dev-harbor.personalhealthtrain.de/demo-station-STATION_NUM/TRAIN_ID",
    "tag": "latest",
    "volumes":
        {"/home/ubuntu/station/station_data/ds_STATION_NUM":
        {"bind":
            "/opt/train_data/ds",
            "mode": "ro"}},
    "gpus":"all"
}

Feedback

Please use the link to answer shortly a few question: Survey