Trust, demonstrated · real execution

Verified setup: we actually ran it

Anyone can copy-paste setup commands. This page shows the real output of the official check_environment.py executed on this workstation — the PASSes, the FAILs, and the Windows bug we hit, all verbatim.

The real run (2026-08-31)

We fetched the script from the repository and ran it with Python 3.11.9:

GR00T-WholeBodyControl Environment Check
Platform: Windows AMD64
Python:   C:\Users\Administrator\AppData\Local\Programs\Python\Python311\python.exe

Basic:
  [+] Python version: 3.11.9
  [+] Git LFS: installed
  [X] CUDA: PyTorch not installed
  [X] PyTorch: not installed (pip install torch)

… crashed at check_disk_space(): AttributeError: module 'os' has no attribute 'statvfs'
What this proves: the script works as designed — it flags missing GPU dependencies (CUDA/PyTorch) before you waste hours. It also exposes a genuine upstream bug: os.statvfs is Unix-only, so the disk-space check crashes on Windows. The basic checks print first, but check_disk_space aborts the run.

Why we publish this

This site's promise is that every fact traces to a source. For setup guidance that means showing the actual environment check, failures included. On a Linux box with CUDA the same script would print the full training/deploy dependency matrix; on this Windows workstation it fails loudly and honestly — which is more useful than a fabricated green checkmark.

Next steps for anyone on Windows: the official deployment guide targets Ubuntu/Debian; for training, Isaac Lab itself is Linux-only in practice. Use WSL2 or a Linux machine — and expect the statvfs crash until upstream fixes it.

Verified setup FAQ

Is this page really showing your own run?
Yes. The log below was produced by fetching the official check_environment.py from the repository and executing it locally on 2026-08-31. PASS/FAIL lines are verbatim from the script's output; nothing is simulated.
Why does the check fail on CUDA / PyTorch?
Because this machine is a Windows workstation without a CUDA-capable GPU configured for PyTorch. That is exactly what the script is for: it tells you what a machine is missing before you try training or deployment. The Python and Git LFS checks pass; the GPU stack does not.
The script crashed on Windows — is that a bug?
Yes, a real one: check_environment.py calls os.statvfs in check_disk_space, which is Unix-only. On Windows it raises AttributeError. We're reporting it upstream; for now the basic checks (Python, LFS) still run and print before the crash.