Isaac Orbit on Triton HPC

Last updated on Dec 16, 2023 2 min read

Isaac Orbit on Triton

Building Isaac Orbit Docker Image

Follow the instructions on setting up docker and nvidia docker images on local machine from the following link.

Create the following dockerfile

Ref: https://github.com/NVIDIA-Omniverse/Orbit/discussions/14

ARG ISAACSIM_VERSION=2022.2.1
FROM nvcr.io/nvidia/isaac-sim:${ISAACSIM_VERSION}

### Use bash by default
SHELL ["/bin/bash", "-c"]

### Install Orbit
RUN apt-get update && \
    DEBIAN_FRONTEND=noninteractive apt-get install -yq --no-install-recommends \
        cmake \
        ncurses-term \
        git && \
    rm -rf /var/lib/apt/lists/* && \
    git clone https://github.com/NVIDIA-Omniverse/Orbit.git --depth 1 -b main /orbit && \
    ln -s "/isaac-sim" /orbit/_isaac_sim && \
    /orbit/orbit.sh --install && \
    /orbit/orbit.sh --extra || exit 0

### Set entrypoint (adjust to your preferences)
ENTRYPOINT ["/orbit/orbit.sh", "-p"]

A docker image with above configurations is built and can be downloaded using the command below.

If you are planning to create your own environments, I used the following work around might not be an efficient way but works.

Copy the /orbit folder from docker on to your triton workspace. If you have created new environments copy them using to the contrib folder using the following command. (Read more about docker cp and scp).

cp -rf ${WRKDIR}/summer-project/Orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/<env_folder_name> ${WRKDIR}/orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/

cd $WRKDIR
singularity build orbit-isaac-2022-2-1.sif docker://harshavguda/orbit-isaacsim:2022.2.1
singularity build orbit-isaac-2022-2-1.sif <name_of_container>

Copy the created .sif file to triton.

To open shell of the docker image run the following command.

srun singularity shell -B /m:/m -B /l:/l -B /scratch:/scratch -B /etc/vulkan/icd.d/nvidia_icd.json -B /etc/vulkan/implicit_layer.d/nvidia_layers.json -B /usr/share/glvnd/egl_vendor.d/10_nvidia.json -B ${WRKDIR}/isaac-sim/kit/cache/Kit:/isaac-sim/kit/cache/Kit,${WRKDIR}/isaac-sim/cache/ov:/isaac-sim/cache/ov,${WRKDIR}/isaac-sim/cache/pip:/isaac-sim/cache/pip,${WRKDIR}/isaac-sim/cache/glcache:/isaac-sim/cache/glcache,${WRKDIR}/isaac-sim/cache/computecache:/isaac-sim/cache/computecache,${WRKDIR}/isaac-sim/logs:/isaac-sim/logs,${WRKDIR}/isaac-sim/data:/isaac-sim/data,${WRKDIR}/isaac-sim/documents:/isaac-sim/documents --nv ${WRKDIR}/orbit-isaac-2022-2-1.sif

Examples

Example script run a custom environment.

#!/bin/bash -l
#SBATCH --gres=gpu:1
#SBATCH --time=02:00:50
#SBATCH --mem=16G
#SBATCH --cpus-per-task=6
#SBATCH --constraint=ampere
export ACCEPT_EULA=Y
export ISAACSIM_PATH="/isaac-sim"
export ISAACSIM_PYTHON_EXE="/isaac-sim/python.sh"
cp -rf ${WRKDIR}/summer-project/Orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/bagmanipulation/ ${WRKDIR}/orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/
cp -rf ${WRKDIR}/summer-project/Orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/bagopening/ ${WRKDIR}/orbit/source/extensions/omni.isaac.contrib_envs/omni/isaac/contrib_envs/

# Run bag lift experiment
srun  singularity exec -B /m:/m -B /l:/l -B /scratch:/scratch -B /etc/vulkan/icd.d/nvidia_icd.json -B /etc/vulkan/implicit_layer.d/nvidia_layers.json -B /usr/share/glvnd/egl_vendor.d/10_nvidia.json -B /scratch/work/gudah1/isaac-sim/kit/cache/Kit:/isaac-sim/kit/cache/Kit,/scratch/work/gudah1/isaac-sim/cache/ov:/isaac-sim/cache/ov,/scratch/work/gudah1/isaac-sim/cache/pip:/isaac-sim/cache/pip,/scratch/work/gudah1/isaac-sim/cache/glcache:/isaac-sim/cache/glcache,/scratch/work/gudah1/isaac-sim/cache/computecache:/isaac-sim/cache/computecache,/scratch/work/gudah1/isaac-sim/logs:/isaac-sim/logs,/scratch/work/gudah1/isaac-sim/data:/isaac-sim/data,/scratch/work/gudah1/isaac-sim/documents:/isaac-sim/documents --nv ${WRKDIR}/orbit-isaac-2022-2-1.sif <exec_command>
Harsha Guda
Harsha Guda

My research interests include reinforcement learning, computer vision and robotics.