Please follow the instructions to install the conda environments and the dependencies of the codebase. We recommend using CUDA 11.x during installations to avoid compatibility issues (remember to replace 11.x
in the following commands with your own CUDA version like 11.4
).
-
Create a new conda environment and activate the environment.
conda create -n mba python=3.8 conda activate mba
-
Install necessary dependencies.
conda install cudatoolkit=11.x pip install -r requirements.txt
-
Install MinkowskiEngine manually following the official instsallation instructions.
mkdir dependencies && cd dependencies conda install openblas-devel -c anaconda export CUDA_HOME=/usr/local/cuda-11.x git clone [email protected]:NVIDIA/MinkowskiEngine.git cd MinkowskiEngine python setup.py install --blas_include_dirs=${CONDA_PREFIX}/include --blas=openblas cd ../..
For CUDA 12.1, the installation of MinkowskiEngine may be incompatible. You may need to add some headers in MinkowskiEngine. Refer to issue#543.
In addition, for some necessary environment variable settings and to deal with module missing problems during the installation process, we recommend you to refer to the MinkowskiEngine section of this blog. -
Install Pytorch3D manually.
cd dependencies git clone [email protected]:facebookresearch/pytorch3d.git cd pytorch3d pip install -e . cd ../..
Hardwares.
- Flexiv Rizon 4 Robotic Arm (or other robotic arms)
- Dahuan AG-95 Gripper (or other grippers)
- Intel RealSense RGB-D Camera (D415/D435/L515)
Softwares.
- Ubuntu 20.04 (tested) with previous environment installed.
- If you are using Flexiv Rizon robotic arm, install the Flexiv RDK to allow the remote control of the arm. Specifically, download FlexivRDK v0.9 and copy
lib_py/flexivrdk.cpython-38-[arch].so
to thedevice/robot/
directory. Please specify[arch]
according to your settings. For our platform,[arch]
isx86_64-linux-gnu
. - If you are using Dahuan AG-95 gripper, install the following python packages for communications.
pip install pyserial==3.5 modbus_tk==1.1.3
- If you are using Intel RealSense RGB-D camera, install the python wrapper
pyrealsense2
oflibrealsense
according to the official installation instructions.