ManiSkill 3

Project Installation Guide

This guide provides step-by-step instructions to set up the environment for this project, which is built upon ManiSkill3. The installation is tailored for a system with CUDA 12.4 and PyTorch 2.5.1.

Prerequisites

• NVIDIA GPU with CUDA 12.4 compatible drivers.
• Conda package manager installed.

Installation Steps

1. Create and Activate Conda Environment

First, create a dedicated Conda environment for this project using Python 3.10. Then, activate the newly created environment.

# Create the environment named 'icra2026'
conda create -n icra2026 python=3.10

# Activate the environment
conda activate icra2026

2. Install PyTorch with CUDA 12.4

Install the correct version of PyTorch and its related libraries directly from the official pytorch and nvidia conda channels. This ensures that the GPU-accelerated libraries are linked correctly.

conda install pytorch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 pytorch-cuda=12.4 -c pytorch -c nvidia

3. Install Python Dependencies

Install all the required Python packages using the requirements.txt file. This file is configured to download packages compatible with our specific PyTorch and CUDA versions.

pip install -r requirements.txt

4. Install Custom ManiSkill Version

Finally, install the version of ManiSkill included in this repository in "editable" mode. The -e flag links the installation to this source directory, so any changes you make to the code are immediately effective in your environment.

pip install -e .

After this step, your environment is fully configured and ready to use.

---

Verifying the Installation

To ensure PyTorch can correctly see and use your GPU, you can run the following Python code:

import torch

if torch.cuda.is_available():
    print(f"✅ Success! PyTorch can see your GPU.")
    print(f"Device Name: {torch.cuda.get_device_name(0)}")
else:
    print("❌ Failure. PyTorch cannot see your GPU.")

---

Citation

If you use ManiSkill3 (versions mani_skill>=3.0.0) in your work, please cite the ManiSkill3 paper:

@article{taomaniskill3,
  title={ManiSkill3: GPU Parallelized Robotics Simulation and Rendering for Generalizable Embodied AI},
  author={Stone Tao and Fanbo Xiang and Arth Shukla and Yuzhe Qin and Xander Hinrichsen and Xiaodi Yuan and Chen Bao and Xinsong Lin and Yulin Liu and Tse-kai Chan and Yuan Gao and Xuanlin Li and Tongzhou Mu and Nan Xiao and Arnav Gurha and Viswesh Nagaswamy Rajesh and Yong Woo Choi and Yen-Ru Chen and Zhiao Huang and Roberto Calandra and Rui Chen and Shan Luo and Hao Su},
  journal = {Robotics: Science and Systems},
  year={2025},
}

License

All rigid body environments in ManiSkill are licensed under fully permissive licenses (e.g., Apache-2.0).

The assets are licensed under CC BY-NC 4.0.