1. Environment

Python=3.7 Pysyft=0.2.x torch=1.4.0 ...

2. Installation

• Install Pysyft 0.2.x-fix-training package:

wget https://github.com/OpenMined/PySyft/archive/refs/heads/ryffel/0.2.x-fix-training.zip
unzip 0.2.x-fix-training.zip

• Install conda:
  1. miniconda
  2. anaconda
• Create conda env:

conda create -n <env_name> python=3.7
conda activate <env_name>

• Install dependency:

cd 0.2.x-fix-training
pip install -e .

If you encountered some errors when installing dependency, try to comment out dict object extras_require in setup.py. Maybe pip install protobuf~=3.19.0 is also required. And when you fail to install pylibsrtp, try apt install libsrtp2-dev.

• Pysyft 0.2.x bug fix: Because 0.2.x version is not supported anymore, you may encounter some bug when using it. If you want to run this project correctly, try to modify 0.2.x-fix-training/syft/frameworks/torch/tensors/interpreters/gradients.py line 40:

<<< old code 
if not isinstance(self.other.child, int):
>>> new code
if torch.is_tensor(self.other) and not isinstance(self.other.child, int):

3. Backend Deployment

• update the Flask (from 1.1.1 to 2.2.5) warning: syft 0.2.9 is not supported by flask 2.2.5

pip uninstall flask
pip install flask