Skip to content

install.md

Latest commit

 

History

History
67 lines (45 loc) · 1.94 KB

install.md

File metadata and controls

67 lines (45 loc) · 1.94 KB

Step-by-step installation instructions

VoxDet is developed based on the official CGFormer codebase and the installation follows similar steps.

a. Create a conda virtual environment and activate

python 3.8 may not be supported.

conda create -n voxdet python=3.7 -y
conda activate voxdet

b. Install PyTorch and torchvision following the official instructions

# Using pip is recommended

pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu113/torch_stable.html

or

conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch -c conda-forge

We selected this PyTorch version because mmdet3d 0.17.1 does not support PyTorch ≥ 1.11.

Note: The following pip installation requires CUDA 11.X on your local machine.

  • To avoid compilation issues, ensure your CUDA toolkit version matches the CUDA version used to compile PyTorch (e.g., 11.x).
  • Use nvcc -V to verify your CUDA version.
  • After compilation, you can use both CUDA 11.X and 12.X for training.

c. Install mmcv, mmdet, and mmseg

pip install openmim
mim install mmcv-full==1.4.0
mim install mmdet==2.14.0
mim install mmsegmentation==0.14.1

c. Install mmdet3d 0.17.1 and DFA3D

Compared with the offical version, the mmdetection3d provided by OccFormer further includes operations like bev-pooling, voxel pooling. After this step, make sure mmdet3d appearing on your pip list.

cd packages
bash setup.sh
cd ../

d. Install other dependencies, like timm, einops, torchmetrics, spconv, pytorch-lightning, etc.

pip install -r docs/requirements.txt

e. Fix bugs (known now)

pip install yapf==0.40.0
pip3 install natten==0.14.6+torch1101cu113 -f https://shi-labs.com/natten/wheels