Support Matrix#

This matrix describes the supported AutoData v0.1.0 stack. Upstream projects may support additional platforms, but configurations outside this matrix are not validated by AutoData. Use the recursively pinned submodules from the AutoData repository instead of independently selecting Isaac Lab or Isaac Lab-Arena revisions.

AutoData platform and resource support#

Area

Supported or required configuration

Host operating system

Linux x86_64; Ubuntu 22.04 or 24.04

GPU

NVIDIA RTX GPU with RT cores and at least 16 GB VRAM. See Isaac Sim requirements.

NVIDIA driver

NVIDIA lists Linux 595.58.03 as a tested driver. See the current Isaac Sim requirements.

CPU and RAM

Minimum: 4 CPU cores and 32 GB RAM. Recommended: 8 or more cores and 64 GB RAM. See Isaac Sim requirements.

Disk

Minimum: 100 GB SSD. Recommended: 500 GB SSD

Docker installation

Docker Engine with the NVIDIA Container Toolkit; recommended AutoData installation

Conda installation

Optional: Python 3.12, conda, uv >= 0.12.21, CMake, and a C++ compiler; conda_installer.sh installs Arena’s locked environment. See Installation.

Isaac Sim

nvcr.io/nvidia/isaac-sim:6.1.0 (Docker) or isaacsim==6.1.0.0 (conda/pip)

Isaac Lab

3.0.0 at commit 28a386f063e41c04c07f50e63eefb83fd8408fbe

Isaac Lab-Arena

0.3.1 at commit 481f5ae5f19df7bc5d24a17240fc0d96ac81eaf7

Python and PyTorch

Python 3.12 and PyTorch 2.11.0 with CUDA 12.8

NumPy, Warp, and Newton

NumPy 2.3.1, Warp 1.16.0, and Newton 1.5.2; see runtime-constraints.txt

Git and Git LFS

Git with recursive submodule support and Git LFS

cuRobo and SkillGen

Optional; cuRobo commit ebb71702f3f70e767f40fd8e050674af0288abe8 and CUDA Toolkit 12.8