- Dockerfile for reproducible
PyVertical
environment - partitioned dataset extending
syft
'sBaseDataset
for remote worker functionality VerticalDataset
class for storing multiple partitioned datsets on remote workers- Changelog
- Moved dataset splitting helper functions to "utils" folder
- Pinned
syft
version in requirements to< 0.3.0
- Install
openmined_psi
pip package to get PSI code
- Removed bazel dependency and
WORKSPACE
file
0.1.0 - 2020-07-10
PoC release of PyVertical
-
demonstrating the concept on MNIST.
- Helper functions for vertically partitioning MNIST dataset into datasets of images and labels
- Helper functions to randomly remove and shuffle datapoints in a dataset
- Dataloader class for loading images and labels from separate datasets
- PSI function to locally match datapoint from separate datasets
- Single-headed split neural network
- Notebook demonstrating local vertical federated learning on MNIST
- Code tested on Ubuntu 18.04, Python 3.6, 3.7, 3.8