Windows version is not working

[pavlos-p]

Description

PyVertical is currently experiencing issues with the Windows version. Syft version 0.2.* requires PyTorch version 1.4.0 and Torchvision version 0.5.0. However, PyTorch version 1.4.0 gives an ImportError: DLL load failed: The operating system cannot run %1.. This bug is known on the PyTorch community (https://pytorch.org/docs/stable/notes/windows.html) with clarification:

This is actually an upstream issue of Anaconda. When you initialize your environment with conda-forge channel, this issue will emerge. You may fix the intel-openmp libraries through this command.

conda install -c defaults intel-openmp -f

The command above does not solve the issue at least for me.

An update to PyTorch version 1.6.0 (conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=9.2 -c pytorch) seems to solve this issue even if Syft version 0.2.9 is not supported.

This is probably a bug that will be solved on updating Syft version to 0.3 and PyTorch version to 1.6 or 1.7.

Blocked by Syft version 0.2.* and PyTorch version 1.4.0.

How to Reproduce

1. Create the environment on Windows operating system (conda env create -f environment.yml).
2. Activate the environment (conda activate pyvertical-dev).
3. Install the Jupyter Notebook (conda install notebook) (If you will not do that, jupyter notebook will still be activated but it will use Base environment instead of pyvertical-dev, and you will get an error ModuleNotFoundError: No module named 'torch'.)
4. Import PyTorch (import torch)
5. See error ImportError: DLL load failed: The operating system cannot run %1..

Expected Behavior

Import PyTorch without any issue.

System Information

• OS: Windows 10
• OS Version: 20H2
• OS Build: 19042.630
• Language Version: [e.g. Python 3.8]
• Package Manager Version: [Anaconda channels: conda-forge, pytorch]