FROM nvidia/cuda:11.1.1-cudnn8-devel-ubuntu20.04 ENV DEBIAN_FRONTEND=noninteractive ENV TZ=Europe/Stockholm RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone ENV NVIDIA_DRIVER_CAPABILITIES ${NVIDIA_DRIVER_CAPABILITIES},compute,display SHELL [ "/bin/bash", "--login", "-c" ] # To fix GPG key error when running apt-get update RUN apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/3bf863cc.pub RUN apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/7fa2af80.pub #install libs first RUN apt-get update -q && \ apt-get install -q -y \ wget \ python3.8-dev \ python3-pip \ python3.8-tk \ git \ ninja-build \ ffmpeg libsm6 libxext6 libglib2.0-0 libsm6 libxrender-dev libxext6 \ && apt-get clean \ && rm -rf /var/lib/apt/lists/* # intall pytorch ENV TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0 7.5 8.0+PTX" ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all" ENV PATH=${PATH}:/usr/local/cuda:/usr/local/cuda/bin RUN pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html RUN pip uninstall numpy scikit-image pandas matplotlib shapely setuptools urllib3 -y RUN pip install numpy==1.20.0 scikit-image==0.19.3 pandas==1.4.4 matplotlib==3.6 shapely==1.8.5.post1 setuptools==59.5.0 RUN pip install scikit-learn pyquaternion cachetools descartes future tensorboard RUN pip install IPython # Install MMCV-series ENV CUDA_HOME=/usr/local/cuda ENV FORCE_CUDA="1" RUN pip install mmcv-full==1.4.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html RUN pip install mmdet==2.14.0 RUN pip install mmsegmentation==0.14.1 # Install mmdetection3d RUN git clone --branch v0.17.1 --single-branch https://github.com/open-mmlab/mmdetection3d.git WORKDIR /mmdetection3d RUN pip install scipy==1.7.3 scikit-image==0.20.0 RUN pip install --no-cache-dir -v -e . # Copy to python dist-package foldder RUN cp -r /mmdetection3d/mmdet3d /usr/local/lib/python3.8/dist-packages # Install UniAD from source WORKDIR / RUN git clone https://github.com/OpenDriveLab/UniAD.git WORKDIR /UniAD RUN pip install -r requirements.txt # We need this old version of torchmetrics to run UniAD... RUN pip install torchmetrics==0.8.2 # Link python to python3 RUN ln /usr/bin/python3 /usr/bin/python