Docker

GRiD does not currently ship an official Docker image. The repo’s two install scripts (install/base_install.sh and install/developer_install.sh) are the supported install path on bare-metal hosts.

If you want to run GRiD in a container, the recipe below is a starting point that mirrors what the install scripts do. It is not tested in CI — adapt to your CUDA version and host arch before relying on it.

Reference Dockerfile sketch

# Pick a CUDA base matching your target compute capability. The repo
# currently targets sm_120 (RTX 5090) and CUDA 13.x; older sm_8x
# GPUs work with CUDA 12.x. Use a -devel image so nvcc + cuda headers
# are present.
FROM nvidia/cuda:13.2.0-devel-ubuntu24.04

# System build deps used by install/developer_install.sh for the Pinocchio
# pybind11 extension. Plus python and the standard build tools.
RUN apt-get update && apt-get install -y --no-install-recommends \
        python3 python3-venv python3-pip \
        git pkg-config g++ \
        libeigen3-dev liburdfdom-headers-dev \
    && rm -rf /var/lib/apt/lists/*

# Clone the repo (or COPY a local checkout in instead).
WORKDIR /opt
RUN git clone --recurse-submodules https://github.com/A2R-Lab/GRiD.git
WORKDIR /opt/GRiD

# End-user install (creates the .venv used by all scripts).
RUN bash install/base_install.sh

# Developer install (adds Pinocchio + robot_descriptions + Pinocchio
# second-order pybind11 extension). Comment out if you only need the
# codegen CLI.
RUN bash install/developer_install.sh

# Make the CLI available on PATH.
ENV PATH="/opt/GRiD/.venv/bin:${PATH}"

CMD ["bash"]

Running with GPU access

Use the NVIDIA Container Toolkit to expose the host GPU:

docker build -t grid:dev .
docker run --rm -it --gpus all grid:dev

Caveats

• The image will be large (CUDA devel + Eigen + Pinocchio is several GB).

This recipe is a known-incomplete starting point; an officially supported Docker image is on the long-term wishlist.