SwiftSyft makes it easy for you to train and inference PySyft models on iOS devices. This allows you to utilize training data located directly on the device itself, bypassing the need to send a user's data to a central server. This is known as federated learning.
- βοΈ Training and inference of any PySyft model written in PyTorch or TensorFlow
- π€ Allows all data to stay on the user's device
- π Support for delegation to background task scheduler
- π Support for JWT authentication to protect models from Sybil attacks
- π Host of inbuilt best practices to prevent apps from over using device resources.
- π Charge detection to allow background training only when device is connected to charger
- π€ Sleep and wake detection so that the app does not occupy resource when user starts using the device
- πΈ Wifi and metered network detection to ensure the model updates do not use all the available data quota
- π All of these smart defaults are easily are overridable
- :mortarboard: Support for both reactive and callback patterns so you have your freedom of choice (_in progress)
- π Support for secure multi-party computation and secure aggregation protocols using peer-to-peer WebRTC connections (in progress).
There are a variety of additional privacy-preserving protections that may be applied, including differential privacy, muliti-party computation, and secure aggregation.
OpenMined set out to build the world's first open-source ecosystem for federated learning on web and mobile. SwiftSyft is a part of this ecosystem, responsible for bringing secure federated learning to iOS devices. You may also train models on Android devices using KotlinSyft or in web browsers using syft.js.
If you want to know how scalable federated systems are built, Towards Federated Learning at Scale is a fantastic introduction!
Cocoapods is a dependency manager for Cocoa projects. Just add OpenMinedSwiftSyft
to your Podfile like below:
pod 'OpenMinedSwiftSyft', ~> 0.1.3-beta1
As a developer, there are few steps to building your own secure federated learning system upon the OpenMined infrastructure:
- π€ Generate your secure ML model using PySyft. By design, PySyft is built upon PyTorch and TensorFlow so you don't need to learn a new ML framework. You will also need to write a training plan (training code the worker runs) and an averaging plan (code that PyGrid runs to average the model diff).
- π Host your model and plans on PyGrid which will deal with all the federated learning components of your pipeline. You will need to set up a PyGrid server somewhere, please see their installation instructions on how to do this.
- π Start training on the device!
π The entire workflow and process is described in greater detail in our project roadmap.
You can use SwiftSyft as a front-end or as a background service. The following is a quick start example usage:
// This is a demonstration of how to use SwiftSyft with PyGrid to train a plan on local data on an iOS device
// Authentication token
let authToken = /* Get auth token from somewhere (if auth is required): */
// Create a client with a PyGrid server URL
if let syftClient = SyftClient(url: URL(string: "ws://127.0.0.1:5000")!, authToken: authToken) {
// Store the client as a property so it doesn't get deallocated during training.
self.syftClient = syftClient
// Create a new federated learning job with the model name and version
self.syftJob = syftClient.newJob(modelName: "mnist", version: "1.0.0")
// This function is called when SwiftSyft has downloaded the plans and model parameters from PyGrid
// You are ready to train your model on your data
// modelParams - Contains the tensor parameters of your model. Update these tensors during training
// and generate the diff at the end of your training run.
// plans - contains all the torchscript plans to be executed on your data.
// clientConfig - contains the configuration for the training cycle (batchSize, learning rate) and metadata for the model (name, version)
// modelReport - Used as a completion block and reports the diffs to PyGrid.
self.syftJob?.onReady(execute: { modelParams, plans, clientConfig, modelReport in
// This returns an array for each MNIST image and the corresponding label as PyTorch tensor
// It divides the training data and the label by batches
guard let MNISTDataAndLabelTensors = try? MNISTLoader.loadAsTensors(setType: .train) else {
return
}
// This loads the MNIST tensor into a dataloader to use for iterating during training
let dataLoader = MultiTensorDataLoader(dataset: MNISTDataAndLabelTensors, shuffle: true, batchSize: 64)
// Iterate through each batch of MNIST data and label
for batchedTensors in dataLoader {
// We need to create an autorelease pool to release the training data from memory after each loop
autoreleasepool {
// Preprocess MNIST data by flattening all of the MNIST batch data as a single array
let MNISTTensors = batchedTensors[0].reshape([-1, 784])
// Preprocess the label ( 0 to 9 ) by creating one-hot features and then flattening the entire thing
let labels = batchedTensors[1]
// Add batch_size, learning_rate and model_params as tensors
let batchSize = [UInt32(clientConfig.batchSize)]
let learningRate = [clientConfig.learningRate]
guard
let batchSizeTensor = TorchTensor.new(array: batchSize, size: [1]),
let learningRateTensor = TorchTensor.new(array: learningRate, size: [1]) ,
let modelParamTensors = modelParams.paramTensorsForTraining else
{
return
}
// Execute the torchscript plan with the training data, validation data, batch size, learning rate and model params
let result = plans["training_plan"]?.forward([TorchIValue.new(with: MNISTTensors),
TorchIValue.new(with: labels),
TorchIValue.new(with: batchSizeTensor),
TorchIValue.new(with: learningRateTensor),
TorchIValue.new(withTensorList: modelParamTensors)])
// Example returns a list of tensors in the folowing order: loss, accuracy, model param 1,
// model param 2, model param 3, model param 4
guard let tensorResults = result?.tupleToTensorList() else {
return
}
let lossTensor = tensorResults[0]
lossTensor.print()
let loss = lossTensor.item()
let accuracyTensor = tensorResults[1]
accuracyTensor.print()
// Get updated param tensors and update them in param tensors holder
let param1 = tensorResults[2]
let param2 = tensorResults[3]
let param3 = tensorResults[4]
let param4 = tensorResults[5]
modelParams.paramTensorsForTraining = [param1, param2, param3, param4]
}
}
// Generate diff data and report the final diffs as
let diffStateData = try plan.generateDiffData()
modelReport(diffStateData)
})
// This is the error handler for any job exeuction errors like connecting to PyGrid
self.syftJob?.onError(execute: { error in
print(error)
})
// This is the error handler for being rejected in a cycle. You can retry again
// after the suggested timeout.
self.syftJob?.onRejected(execute: { timeout in
if let timeout = timeout {
// Retry again after timeout
print(timeout)
}
})
// Start the job. You can set that the job should only execute if the device is being charge and there is
// a WiFi connection. These options are on by default if you don't specify them.
self.syftJob?.start(chargeDetection: true, wifiDetection: true)
}
See API Documenation for complete reference.
A mini tutorial on how to run SwiftSyft
on iOS using the background task scheduler can be found here
The demo app fetches the plans, protocols and model weights from PyGrid server hosted locally. The plans are then deserialized and executed using libtorch.
Follow these steps to setup an environment to run the demo app:
- Clone the repo PyGrid and change directory to
PyGrid/apps/domain
$ git clone https://github.com/OpenMined/PyGrid
$ cd PyGrid/apps/domain
- Run the PyGrid Domain application
$ ./run.sh --port 5000 --start_local_db
- Install PySyft from source or via PyPy. Follow the instructions specified in the repo.
- Clone the
PySyft
repo. In your command line, go toPySyft/examples/federated-learning/model-centric/
folder and run jupyter notebook.
$ cd PySyft/examples/federated-learning/model-centric/
$ jupyter notebook
-
Open the notebook
mcfl_create_plan_mobile.ipynb
notebook. Run all the cells to host a plan to your running PyGrid domain server. -
Set-up demo project using Cocoapods
-
Install Cocoapods
gem install cocoapods
- Install the dependencies of the project.
pod install # On the root directory of this project
- Open the file
SwiftSyft.xcworkspace
in Xcode. - Run the
SwiftSyft
project. It automatically uses127.0.0.1:5000
as the PyGrid URL.
You can work on the project by running pod install
in the root directory. Then open the file SwiftSyft.xcworkspace
in Xcode. When the project is open on Xcode, you can work on the SwiftSyft
pod itself in Pods/Development Pods/SwiftSyft/Classes/*
- Star, for and clone the repo
- Open the project in Xcode
- Check out current issues and Github. For newcomers, check out issues labeled
Good first issue
- Do your work
- Push your fork
- Submit a PR to OpenMined/SwiftSyft
Read the contribution guide as a good starting place.
For support in using this library, please join the #lib_swift_syft Slack channel. If you'd like to follow along with any code changes to the library, please join #code_swiftsyft Slack channel. Click here to join our Slack Community!
Thanks goes to these wonderful people (emoji key):
Mark Jimenez π» π |
Madalin Mamuleanu π» |
Rohith Pudari π» |
Sebastian Bischoff π |
Luke Reichold π» |
This project follows the all-contributors specification. Contributions of any kind welcome!