Skip to content

Latest commit

 

History

History
86 lines (58 loc) · 2.27 KB

env_install.md

File metadata and controls

86 lines (58 loc) · 2.27 KB
Raw

Environment Configuration Document

CUDA

Confirm cuda-11.8 is installed

ls /usr/local

If cuda-11.8 is not found, go to the NVIDIA Official CUDA Download Page to download CUDA 11.8. Select the appropriate operating system and installation method (usually a .run file or .deb package). For example, with .run:

sudo chmod +x cuda_11.8.*_linux.run # Grant executable permission to the .run file
sudo ./cuda_11.8.*_linux.run # Run the installation program

Configure Environment Variables

After installing in /usr/local, you need to add the CUDA-related paths to the system's environment variables.

  1. Edit the .bashrc file (or .zshrc file if you use Zsh):

    nano ~/.bashrc

  2. Add the following content at the end of the file:

    export PATH=/usr/local/cuda-11.8/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

  3. Save and exit the editor, then load the new environment variables using the following command:

    source ~/.bashrc

Create a Virtual Environment

Assume install virtualenvwrapper first

mkvirtualenv vadv2  --python=python3.8

Set Mirror Source for Best Download Speed

pip config set global.index-url https://mirrors.bfsu.edu.cn/pypi/web/simple

Install Dependency Libraries

Note: Torch requires cuda-11.8

python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
python -m pip install -r preinstall_requirements.txt

Configure Environment Variables

export PATH=/usr/bin:$PATH
export CUDA_HOME=/usr/local/cuda-11.8

Install mmcv Library

Note: It is necessary to compile mmcv on the local machine with cuda-11.8, otherwise it will report an error error in ms_deformable_im2col_cuda: no kernal image is available for execution on the device

python -m pip install -v -e .

Download Pre-trained Models

Note: If ckpts already exists, skip this step

cd ckpts
wget https://hf-mirror.com/rethinklab/Bench2DriveZoo/resolve/main/resnet50-19c8e357.pth
wget https://hf-mirror.com/rethinklab/Bench2DriveZoo/resolve/main/r101_dcn_fcos3d_pretrain.pth