d2go on Window11

Requirement

Python 3.8
CUDA 12.2
Windows 11
Anaconda 3 (Optional)

Preparation

You need to know first, your environment ready for GPU usage.

Let's start from installing the correct torch version (Ref. https://pytorch.org/get-started/locally/) which match with current CUDA version on your Windows, in this case it is 12.2 :

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121

Checking torch with CUDA available or not :

python -c 'import torch;import torchvision; print(torch.cuda.is_available())'

Installation

[Optional] Install python kernel module to support Jupyter notebook :
pip install ipykernel
Create the kernel with python 3.8 :
python -m ipykernel install --user --name py38gpu

Create environment for anaconda :

conda create -n d2go python=3.8.0
conda activate d2go

Install torch:

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu121

Install detectron2's dependencies :

python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'

If you got error like this :

...

      3 errors detected in the compilation of "C:/Users/patha/AppData/Local/Temp/pip-req-build-7pdw3whx/detectron2/layers/csrc/nms_rotated/nms_rotated_cuda.cu".
      nms_rotated_cuda.cu
      error: command 'C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v12.2\\bin\\nvcc.exe' failed with exit code 2
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for detectron2
  Running setup.py clean for detectron2
Failed to build detectron2
ERROR: Could not build wheels for detectron2, which is required to install pyproject.toml-based projects

Please clone the repository :

git clone https://github.com/facebookresearch/detectron2 detectron2_repo

and set FORCE_CUDA="1" in your terminal (ex. PowerShell):

set FORCE_CUDA="1"

Then install the dependencies manually (it requires ninja to build package) :

pip install ninja
pip install -e ./detectron2_repo

Install mobile_cv :

python -m pip install 'git+https://github.com/facebookresearch/mobile-vision.git'

Then d2go :

git clone https://github.com/facebookresearch/d2go d2go_repo
pip install -e ./d2go_repo

Usage

Ref. notebook

Other Issues

Docker : WSL2 got low memory usage

Please try to update your C:\Users\<YOUR_USER_NAME>\.wslconfig :

# Settings apply across all Linux distros running on WSL 2
[wsl2]

# Limits VM memory to use no more than 4 GB, this can be set as whole numbers using GB or MB
memory=8GB 

# Sets the VM to use two virtual processors
processors=2

# Sets amount of swap storage space to 8GB, default is 25% of available RAM
swap=8GB

# Turn off default connection to bind WSL 2 localhost to Windows localhost
localhostforwarding=true

# Turns on output console showing contents of dmesg when opening a WSL 2 distro for debugging
debugConsole=true

patharanordev / d2go