Syft decouples private data from model training, using Federated Learning, Differential Privacy, and Encrypted Computation (like Multi-Party Computation (MPC) and Homomorphic Encryption (HE)) within the main Deep Learning frameworks like PyTorch and TensorFlow. Join the movement on Slack.
Most software libraries let you compute over the information you own and see inside of machines you control. However, this means that you cannot compute on information without first obtaining (at least partial) ownership of that information. It also means that you cannot compute using machines without first obtaining control over those machines. This is very limiting to human collaboration and systematically drives the centralization of data, because you cannot work with a bunch of data without first putting it all in one (central) place.
The Syft ecosystem seeks to change this system, allowing you to write software which can compute over information you do not own on machines you do not have (total) control over. This not only includes servers in the cloud, but also personal desktops, laptops, mobile phones, websites, and edge devices. Wherever your data wants to live in your ownership, the Syft ecosystem exists to help keep it there while allowing it to be used privately for computation.
The current stable release is 0.5.0 which is available on:
For many use cases you can simply use:
$ pip install syft
If you are doing the Private AI Series or you are an external party developing against Syft and Grid please use the syft_0.5.0 branch.
This is the dev branch and to accommodate our need to experiment with new ideas and implementations we will be moving a few things around during the early stages of 0.6.0. Currently the core syft library and code will remain fairly stable, while we do some much needed quality improvements and refactors to the grid codebase and its tooling for deployment and orchestration of nodes.
During the process of development we will be moving examples from the /packages/syft/examples folder down to the /notebooks folder and ensuring they are working and tested with the latest dev code.
This repo contains multiple sub-projects which work together.
OpenMined/PySyft
├── README.md <-- You are here 📌
└── packages
├── grid <-- Grid - A network aware, persistent & containerized node running Syft
├── notebooks <-- Notebook Examples and Tutorials
└── syft <-- Syft - A package for doing remote data science on private data
To read more about what Syft is please consult the current 0.5.0 README.
To read more about what Grid is please consult the old PyGrid README until we finish writing the new one.
- docker
- tox
- python 3.7+
You will need docker and docker-compose to do development on the monorepo tooling.
Run the FastAPI Dev environment using:
$ cd packages/grid
$ source .env && docker compose up
$ cd packages/grid
$ docker compose build
You will need tox to run some of our build and test tools.
$ pip install tox
$ tox -l
You should see the following:
syft.jupyter
syft.lint
syft.test.fast
syft.test.libs
syft.test.duet
syft.test.security
These commands can be run like so:
$ tox -e syft.lint
We are providing a simple way to deploy all of our stack inside a single VM so that no matter where you want to run everything you can do so easily by thinking in terms of a single machine either bare metal or VM and have it provisioned and auto updated.
To develop against this locally you will want the following:
- vagrant
- virtualbox
- ansible
- hagrid <-- in packages/hagrid
You can install HAGrid with pip:
$ pip install "git+https://github.com/OpenMined/PySyft@demo_strike_team_branch_4#subdirectory=packages/hagrid"
$ brew install vagrant virtualbox ansible
Hagrid the Grid deployment tool:
$ cd packages/hagrid
$ pip install -e .
Vagrant allows us to create and manage VMs locally for development. During the startup
process of creating the VM the ansible provisioning scripts will be applied automatically
to the VM. If you change the Vagrantfile which describes how the VM is defined you will
need to either vagrant reload or destroy and re-create it.
Making changes to the VM state should be done through the ansible scripts so that
the state of the box is idempotent and re-running the ansible provisioning scripts
should always result in the same working grid node state.
To allow rapid development we mount the PySyft source repo into the VM at the path:
/home/om/PySyft which is where it would be if it was cloned down on a real remote VM.
The configuration is done via a Vagrantfile which is written in ruby.
The VM will be accessible on the IP 10.0.1.2 which is defined in the Vagrantfile.
The Landrush plugin for vagrant gives us an automatic dns service so we can access our local VM as though it were a real live domain on the internet.
$ vagrant plugin install landrush
With this enabled you can access the box on:
http://node.openmined.grid
NOTE: You may need your sudo password to enable the landrush DNS entry on startup.
$ cd packages/grid
$ vagrant up --provision
You want to do this any time you are testing out your ansible changes.
$ cd packages/grid
$ vagrant provision
If you want to do a quick deploy where you skip the system provisioning you can run:
$ ANSIBLE_ARGS='--extra-vars "deploy_only=true"' vagrant provision
$ cd packages/grid
$ vagrant ssh
Create a VM on your cloud provider with Ubuntu 20.04 with at least:
- 2x CPU
- 4gb RAM
- 40gb HDD
Generate or supply a private key and note down the username.
Run the following:
$ hagrid launch node --type=domain --host=104.42.26.195 --username=ubuntu --key_path=~/.ssh/key.pem
If you want to later skip the setup process of installing packages and docker engine etc you can pass in --mode=deploy which will skip those steps.
If you wish to use a different fork of PySyft you can pass in --repo=The-PET-Lab-at-the-UN-PPTTT/PySyft --branch=ungp_pet_lab
$ sudo su - om
Also, join the rapidly growing community of 12,000+ on Slack. The Slack community is very friendly and great about quickly answering questions about the use and development of PySyft!
This software is in beta. Use at your own risk.
For support in using this library, please join the #support Slack channel. Click here to join our Slack community!
We are very grateful for contributions to Syft and Grid from the following organizations!
 |
 |
 |
 |
 |
 |