This codebase is tested with torch==1.12.1, mmengine==0.10.4, mmcv==2.2.0, mmdet==3.3.0, and mmdet3d==1.4.0, with CUDA 11.3.
Step 1. Create a conda environment and activate it.
conda create --name superflow python==3.8 -y
conda activate superflowStep 2. Install PyTorch following official instructions.
conda install pytorch torchvision -c pytorchStep 3. Install MMEngine, MMCV, MMDetection, MMDetection3D using MIM.
pip install -U openmim
mim install mmengine
mim install mmcv
mim install mmdet
mim install mmdet3dOptionally, you can also install the above projects from the source, e.g.:
git clone https://github.com/open-mmlab/mmdetection3d
cd mmdetection3d
pip install -v -e .Meanwhile, you also need to install nuScenes-devkit.