install-isaacsim

작성자: nvidia

pip 또는 소스 빌드를 통해 Isaac Sim을 설치합니다. Docker 설정, 확인 및 일반적인 설치 문제를 다룹니다. 사용자가 설치, 설정 또는 빌드를 요청할 때 사용하세요.

npx skills add https://github.com/nvidia/omniperf --skill install-isaacsim

Install Isaac Sim

Public repo: https://github.com/isaac-sim/IsaacSim Branch convention: develop (latest), release/* (stable), version tags (e.g., 6.0.0)

System Requirements

  • GPU: NVIDIA RTX (Ada, Ampere, or newer recommended)
  • Driver: 535+ (check with nvidia-smi)
  • OS: Ubuntu 22.04+ (Linux), Windows 10/11
  • Python: see Python version matrix below (version is pinned per Isaac Sim release)
  • RAM: 32 GB+ recommended
  • Disk: ~30 GB for full install with cached assets

Python version matrix

Isaac Sim is built against a single, specific CPython version per release. Using the wrong Python version will cause pip install to fail to resolve wheels, or import-time ABI errors for the source build. Match the table exactly — newer or older minor versions are not supported.

Isaac SimPythonNotes
4.0.x – 4.2.x3.10Linux/Windows
4.5.x3.10Last 3.10 release
5.0.x3.11GLIBC 2.35+ required on Linux
5.1.x3.11
6.0.x3.12Current (Early Developer Release → GA)

Primary sources:

Check your Python:

python3 --version

If the required version isn't installed (Ubuntu example):

sudo add-apt-repository -y ppa:deadsnakes/ppa
sudo apt-get update
sudo apt-get install -y python3.12 python3.12-venv python3.12-dev   # adjust version to match the table

Virtual Environment (Strongly Recommended)

Always install Isaac Sim into an isolated Python environment. Installing into the system Python conflicts with distro packages (especially on Ubuntu) and makes upgrades/uninstalls messy. Use one of the options below before running any pip install command from this skill.

Pick the Python version that matches your target Isaac Sim release — see the Python version matrix. Examples below use 3.12 for Isaac Sim 6.0.x; substitute 3.11 for 5.x or 3.10 for 4.5.x.

Option A: venv (stdlib, simplest)

# Replace 3.12 with the version required by your Isaac Sim release
python3.12 -m venv ~/venvs/isaacsim
source ~/venvs/isaacsim/bin/activate
python -m pip install --upgrade pip setuptools wheel

Deactivate later with deactivate. Re-activate in any new shell before running Isaac Sim commands.

Option B: uv (fast, recommended for CI)

# Install uv once: https://docs.astral.sh/uv/
uv venv --python 3.12 ~/venvs/isaacsim   # match Python to Isaac Sim release
source ~/venvs/isaacsim/bin/activate
uv pip install --upgrade pip

Option C: conda / mamba

conda create -n isaacsim python=3.12 -y   # match Python to Isaac Sim release
conda activate isaacsim

Notes

  • Python version is strict. Each Isaac Sim release is built against one specific CPython minor version — see the matrix. Mismatches cause wheel resolution failures or import-time ABI errors.
  • Method 2 (source build) ships its own Python via _build/linux-x86_64/release/python.sh — you do not need a venv for running the source build, but you still want one for any host-side tooling (tests, scripts, editable installs).
  • Editable install (Method 3) needs an active venv — never pip install -e . into system Python.
  • Cached assets & shader caches live under ~/.cache/ov and ~/.local/share/ov. These are shared across venvs; deleting the venv does not clear them.
  • To completely reset: deactivate && rm -rf ~/venvs/isaacsim ~/.cache/ov ~/.local/share/ov.

Method 1: Pip Install (Quickest)

Note: As of April 2026, NVIDIA's pip install flow uses PyPI plus the NVIDIA package index, installs PyTorch first, and installs the full Isaac Sim extras plus the extension cache. A bare pip install isaacsim installs only the metapackage and can pass an import check while missing app, benchmark, or extension-cache components.

Activate your venv first (see Virtual Environment), then:

# Accept the Omniverse EULA non-interactively for scripts/CI.
export OMNI_KIT_ACCEPT_EULA=YES

# Isaac Sim 6.0 docs use PyTorch 2.10.0; choose the CUDA wheel index for your platform.
pip install torch==2.10.0 --index-url https://download.pytorch.org/whl/cu128

# Latest compatible full install for the active Python.
pip install 'isaacsim[all,extscache]' --extra-index-url https://pypi.nvidia.com

# Or pin to a specific release. Match the Python version matrix above.
# Isaac Sim wheels are published in 4-segment form (e.g. 4.5.0.0, 4.2.0.2);
# pin all four segments so patch releases like 4.2.0.2 are matched explicitly.
pip install 'isaacsim[all,extscache]==4.5.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.10
pip install 'isaacsim[all,extscache]==5.1.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.11
pip install 'isaacsim[all,extscache]==6.0.0.0' --extra-index-url https://pypi.nvidia.com   # Python 3.12

Verify:

python -c "import isaacsim; print('OK')"
python -c "import isaacsim; from isaacsim.simulation_app import SimulationApp; print('SimulationApp OK')"

Method 2: Source Build from GitHub

Use this when you need to modify code, run tests, or link a custom Kit build.

Prerequisites

Docker is required by default. Check and install if missing:

docker ps > /dev/null 2>&1 && echo "DOCKER OK" || echo "NEED INSTALL"

# Install Docker (Ubuntu/Debian)
sudo apt-get update && sudo apt-get install -y docker.io
sudo systemctl start docker && sudo systemctl enable docker
sudo usermod -aG docker $USER

# Start a new login session, or run this shell-only refresh:
newgrp docker

# If docker ps still fails, use `sudo docker ...` for the current command
# or ask an administrator to fix Docker group/session setup.
# Do NOT chmod /var/run/docker.sock: it grants root-equivalent Docker access
# to every local user.

Build

git clone https://github.com/isaac-sim/IsaacSim.git
cd IsaacSim
git checkout develop   # or a specific version tag

# With Docker (default, recommended)
./build.sh -xr

# Without Docker (fallback — needs extra deps)
sudo apt-get install -y libglu1-mesa-dev libegl1-mesa-dev libgles2-mesa-dev patchelf
./build.sh -xr --no-docker

Build output: _build/linux-x86_64/release

Verify source build

cd _build/linux-x86_64/release
./python.sh -c "import omni; print('OK')"

Method 3: Pip Editable Install from Source

Activate your venv first (see Virtual Environment), then:

git clone https://github.com/isaac-sim/IsaacSim.git
cd IsaacSim
git checkout develop
pip install -e .

Nucleus Authentication (Optional)

Only needed for benchmarks/scenes that use Nucleus-hosted assets. Many workflows use local assets and skip this.

export OMNI_USER='$omni-api-token'
export OMNI_PASS='<YOUR-API-TOKEN>'
  • OMNI_USER is always the literal string $omni-api-token (not a shell variable)
  • Ask the user for their OMNI_PASS token if needed
  • Expired JWT tokens cause silent failures (scenes load but materials are missing — all-black renders)

Check token expiry:

echo "$OMNI_PASS" | cut -d. -f2 | base64 -d 2>/dev/null | python3 -c "
import json, sys; from datetime import datetime; d = json.load(sys.stdin)
print(f'Expires: {datetime.fromtimestamp(d.get(\"exp\", 0))}')"

Common Issues

Build fails with Docker permission error

sudo usermod -aG docker $USER
newgrp docker
# If still fails in this shell: use `sudo docker ps` to verify Docker works,
# then start a fresh login session so group membership is applied.

No DISPLAY / Vulkan init fails

Kit needs a display even in headless mode. Set up Xvfb:

sudo apt-get install -y xvfb
Xvfb :99 -screen 0 1920x1080x24 &
export DISPLAY=:99

Shutdown hangs

Do not patch installed Isaac Sim files by default. For repeated Tracy shutdown hangs, use the scoped last-resort workaround in the profiling skill, and restore the original file afterward.


nvidia의 다른 스킬

compileiq-debug
nvidia
Use when something is wrong: Search() hangs, all evaluations return INVALID_SCORE, scores aren't improving, every config returns the same number, ptxas errors…
official
create-github-pr
nvidia
gh CLI를 사용하여 GitHub 풀 리퀘스트를 생성합니다. 사용자가 새 PR을 만들거나, 코드 리뷰를 제출하거나, 풀 리퀘스트를 열고자 할 때 사용합니다. 트리거 키워드 -…
official
diagnose-perf
nvidia
First-responder performance triage for Isaac Sim and Isaac Lab. Identifies bottleneck category (GPU-bound, CPU-bound, VRAM, loading) using nvidia-smi and…
official
eagle3-review-logs
nvidia
Review EAGLE3 pipeline experiment logs from the launcher's experiments/ directory. Summarizes pass/fail status for all 4 tasks, diagnoses failures with root…
official
nemoclaw-maintainer-cross-issue-sweep
nvidia
다른 열린 이슈들을 스캔하여 주어진 PR이 함께 수정하거나 실수로 망가뜨릴 수 있는 이슈를 찾습니다. 인접 수정 기회와 모순 위험을 file:line…과 함께 출력합니다.
official
karpathy-guidelines
nvidia
일반적인 LLM 코딩 실수를 줄이기 위한 행동 지침입니다. 코드 작성, 검토 또는 리팩토링 시 과도한 복잡성을 피하고 정밀한 변경을 위해 사용하세요.
official
fhir-basics
nvidia
에이전트에게 FHIR R4 API의 작동 방식, 사용 가능한 리소스, 검색 매개변수를 사용한 쿼리 방법, 모든 응답 형식을 올바르게 파싱하는 방법을 가르칩니다…
official
underdeclared-agent
nvidia
A helpful assistant agent
official