Environment Setup

EBench uses a client–server architecture. You will set up two environments:

• Server environment — Isaac Sim, cuRobo, and the simulation server code from the GenManip repository.
• Client environment — your model’s own Python environment. The lightweight genmanip-client package is installed alongside your model dependencies and has very few dependencies of its own, so it will not conflict with your model.

Prerequisites

• Linux workstation with an NVIDIA GPU.
• CUDA 12.1 and a compatible driver.
• An Isaac Sim 4.1.0 compatible Python environment (for the server).

Server environment

Clone the simulation server

git clone https://github.com/InternRobotics/GenManip.git
cd GenManip

Install Isaac Sim

export CUDA_HOME=/usr/local/cuda-12.1
pip install isaacsim==4.1.0 isaacsim-extscache-kit==4.1.0 isaacsim-extscache-kit-sdk==4.1.0 isaacsim-extscache-physics==4.1.0 --extra-index-url https://pypi.nvidia.com
pip install torch==2.4.0 --extra-index-url https://download.pytorch.org/whl/cu121

If you have a local Isaac Sim installation, you can use it directly — run server-side commands with /isaac-sim/python.sh instead.

Install project dependencies

mkdir -p saved/envs
git clone https://github.com/NVlabs/curobo.git saved/envs/curobo
pip install -e saved/envs/curobo --no-build-isolation
pip install -r requirements.txt

Client environment

The client lives in the Python environment where your model runs — install it there, alongside your model dependencies. genmanip-client is a standalone repository with minimal dependencies, so it will not conflict with anything in your model environment.

git clone https://github.com/InternRobotics/genmanip-client.git
cd genmanip-client
pip install -e .
gmp --help

Verify

Make sure both environments are working before proceeding:

# Server environment
python ray_eval_server.py --help

# Client environment
gmp --help

Next step: download benchmark assets.