Skip to content

Latest commit

 

History

History
159 lines (119 loc) · 7.58 KB

README.md

File metadata and controls

159 lines (119 loc) · 7.58 KB

Installation Guide for Windows Subsystem for Linux 2

Table of content

Check your Windows version

Windows 11 or Windows 10, version 21H2 is needed to enable GPU acceleration. If you are using Windows 10 version 21H2, you might need to be in the dev channel. I personally tested with Windows 11 and can confirm that it is not need for Windows 11. If you don't have access to any required system, you could try installing WSL2 with CPU only instead.

To check your Windows version:

  • Windows 11: Windows -> Settings -> System -> About
  • Windows 10: Windows -> Settings -> About

To upgrade your system, there are a few options:

Install/Update the NVIDIA driver

Please download and install the appropriate NVIDIA driver on your Windows system. In earlier times, there's a seperate driver for WSL but recently NVIDIA merged it into the currently available drivers. I personally installed the Game Ready Driver(GRD), but I suppose the Studio Drivers also have it.

Note that we will not need to install any drivers in the Linux system.

Install WSL2

If WSL2 is not installed at all, please the following command in your Windows terminal/powershell. This command will download and set up WSL2 with Ubuntu by default

wsl --install

If you want some other Linux distribution, you can run the following

wsl.exe --install -d <Distribution Name>

After the above is done, please go to Windows and search for the distribution that you installed. Take Ubuntu for example, we search for "Ubuntu for Windows". Open it and you should see a line of words saying

Installing, this may take a few minutes ...

Giving some time to install, it should be ready to go in a few minutes. However, you might see something like

The virtual machine could not be started because a required feature is not installed

In this case, there are a few things can be done:

  • Enable Virtual Machine Platform, Windows Subsystem for Linux, Windows Hypervisor Platform:
    Windows -> Search for Turn Windows features on or off -> Look for these features
  • Enable Hardware/CPU virtualization:
    Restart the machine -> go into BIOS mode -> enable virtualization (may have different names depending on you device)

We have successfully installed WSL2 with GPU acceleration by now. To verify that we do have GPU access, we can simply install Pytorch with CUDA and run the following in the python terminal (guide available in the next section)

import torch
torch.cuda.is_available()
torch.cuda.get_device_name(0)

Install softwares

Now you can install your favorite softwares as how you would usually do!

  • Miniconda

Run the following and restart the command shell

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
~/miniconda3/bin/conda init bash
~/miniconda3/bin/conda init zsh

Create and switch to a new environment

conda create --name py39 python=3.9
conda activate py39
  • Pytorch

Go to Pytorch's official website and choose the desired version. For example, to install pytorch 1.11.0 with CUDA 11.3

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
  • CUDA

Go to NVIDIA's official CUDA tool kit archive website, choose the desired release, and follow the instruction there. For example, to install CUDA 11.3.0 on WSL2-Ubuntu, we can do

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.3.0/local_installers/cuda-repo-wsl-ubuntu-11-3-local_11.3.0-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-11-3-local_11.3.0-1_amd64.deb
sudo apt-key add /var/cuda-repo-wsl-ubuntu-11-3-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda-11.3 # different from the official guide, works better when dealing with multiple CUDA releases

First compile CUDA and install Pytorch with the same CUDA version as instructed above. Note that if on UofM Great Lakes cluster, one can simply run the below to load python, openblas, CUDA, and pytorch. Also note that using anaconda on this cluster might be problematic if used in conjunction with pytorch.

module purge
module load python/3.9
module load openblas # workaround for one of the below command if don't have sudo access
module load CUDA/11.3
module load pytorch # Note that this currently loads pytorch 1.12 + CUDA11.6

Let's say we use CUDA 11.3 here. Then run the following

export CXX=gcc
export CUDA_HOME=/usr/local/cuda-11.3
export MAX_JOBS=1 # prevents a weird error
  
# If we use CUDA 10, we need GCC < 8
# export CXX=g++-7
  
sudo apt install build-essential python3-dev libopenblas-dev
pip install ninja
pip install -U MinkowskiEngine --install-option="--blas=openblas" -v --no-deps
  
# For pip installation from the latest source
# pip install -U git+https://github.com/NVIDIA/MinkowskiEngine -v --no-deps \
#                           \ # uncomment the following line if you want to force cuda installation
#                           --install-option="--force_cuda" \
#                           \ # uncomment the following line if you want to force no cuda installation. force_cuda supercedes cpu_only
#                           --install-option="--cpu_only" \
#                           \ # uncomment the following line to override to openblas, atlas, mkl, blas
#                           --install-option="--blas=openblas" \

See this issue if problem persists: NVIDIA/MinkowskiEngine#300

Trouble shooting

  • WSL2 cannot connect to internet

This is usually because the DNS server address is constantly rewritten by WSL2 for some reason. Simply run the following command fixes the problem

sudo rm /etc/resolv.conf
sudo bash -c 'echo "nameserver 8.8.8.8" > /etc/resolv.conf'
sudo bash -c 'echo "[network]" > /etc/wsl.conf'
sudo bash -c 'echo "generateResolvConf = false" >> /etc/wsl.conf'
sudo chattr +i /etc/resolv.conf

Additional Links