Skip to content

Latest commit

 

History

History
65 lines (44 loc) · 3.37 KB

HowTo.md

File metadata and controls

65 lines (44 loc) · 3.37 KB

How-To Guides

Here you can find a collection of recipes for various tasks and settings you might want to try out.

What do you want to do?

1. Dependencies and Environment

  • Create a conda environment and install all dependencies to run our code and reproduce our results:

    • Create a new conda environment with python 3.7.1. Other pythons are not supported!
    • Run make install to recreate the environment
    • Disclaimer: We have only tested our code on Ubuntu 18.04 and 20.04. We cannot guarantee compatibility with any other operating system.

  • Clean up everything and restore the environment to its default state:

    • Run make clean_all

2. Training using VirtualWorkers

  • Simulate training with VirtualWorkers using the Paediatric Pneumonia dataset:

    • Run make symbolic_server_folders to randomly split the Paediatric Pneumonia Dataset into three worker folders and a validation set. You can also run make minimal_server_folders if you are in a real hurry, but results won't be nearly as good.
    • Run make federated_secure to train a model with encrypted federated learning
    • The model weights and a .csv file with metadata will be saved under model_weights

  • Run training on VirtualWorkers using your own data:

    • TODO: add instructions (...) populate folders etc.
    • If you want to use more or fewer workers

3. Training using GridNodes

  • Simulate training with GridNodes using our data:

    • Run run_websocket_server.py ... (finds data automatically)
    • TODO: add instructions for GridNode
    • If you want to use more or fewer workers

  • Run training on real servers using GridNodes and your own data:

    • run websocket server on each cloud instance (finds data automatically)
    • TODO: add instructions
    • If you want to use more or fewer workers

4. CrypTen

  • Use the CrypTen library to perform an inference benchmark on the Paediatric Pneumonia Test Set: WARNING: This is EXTREMELY processor and RAM intense and can result in your machine running out of RAM or locking up.
    • Make sure there is a .pretrained_weights directory in the project root containing the model you want to encrypt
    • Name the model you would like to use crypten_weights.pt
    • Run make crypten_dataset. This will create testdata.pt and testlabels.pt under data/. These correspond to the test set images and labels.
    • Run make crypten_benchmark. This will run a benchmark on two images from the test set and should take about 200 seconds on a powerful machine. It might take much longer depending on your CPU and RAM.

5. Miscellaneous

  • If you want to adapt PriMIA to another use-case altogether, go here

  • Monitor your training with Visdom

    • From the command line, run visdom to start a visdom server. It will be located on localhost:xxxx. Navigate to this page with your browser.
    • For an example, run make federated_secure_visdom which will launch a slightly longer training which you can monitor with Visdom!
    • For more information on Visdom, see here. Scroll down on this page to see configuration options, e.g. port selection or authentication for Visdom and many tutorials!

  • Run MNIST/Run VGG etc. etc.

    • args ...

  • Run a hyperparam optimisation trial (locally)

  • Run a hyperparameter optimisation trial for the whole federation