Installation#

OWRT is a Python package with optional extras. The core package compiles scenes and needs no GPU; Sionna RT, the web service and the tutorial notebooks are opt-in extras.

Requirements#

  • Python 3.10 to 3.12 (the project environment uses 3.11)

  • Conda for the scripted setup, or any virtual environment for a manual install

  • For ray tracing: Sionna RT 2.0.1 or a later 2.0 release, which runs on an NVIDIA GPU through CUDA or on the CPU through the system’s LLVM library (DRJIT_LIBLLVM_PATH names the library when it is not found)

  • Network access to the public geodata providers when a scene is compiled for the first time

Scripted setup with Conda#

The setup script creates the named owrt environment with Python 3.11 and installs the development, Sionna RT and web extras into it:

git clone https://github.com/zhiheng-yang/openworld-radio-twin.git
cd openworld-radio-twin
./scripts/setup.sh
conda activate owrt

Check that Sionna RT imports and which Mitsuba variant it selects:

python -c "import sionna.rt, mitsuba; print(sionna.rt.__version__, mitsuba.variant())"

The variant is cuda_ad_mono_polarized on a visible NVIDIA GPU and llvm_ad_mono_polarized on the CPU. OWRT does not install CUDA or NVIDIA drivers; Sionna uses the runtime already exposed by the machine.

Manual install and extras#

python -m pip install -e .              # scene compiler and dataset generator
python -m pip install -e '.[rt]'        # + Sionna RT 2.0
python -m pip install -e '.[web]'       # + FastAPI service and browser explorer
python -m pip install -e '.[tutorial]'  # + JupyterLab, ipywidgets, pythreejs, Plotly, pandas
python -m pip install -e '.[torch]'     # + PyTorch, for the dataset reader
python -m pip install -e '.[all]'       # everything above
python -m pip install -e '.[dev]'       # pytest and ruff
python -m pip install -e '.[docs]'      # Sphinx toolchain for this documentation

Extra

Adds

Needed for

rt

sionna-rt

owrt.load_scene(...), Sionna dataset radio maps

web

fastapi, uvicorn

owrt serve, the explorer and batch pages, the HTTP API

tutorial

JupyterLab, ipywidgets, pythreejs, Plotly, pandas

The notebooks under Tutorials

torch

PyTorch

RadioMapDataset in openworld_radio_twin.torch_dataset (PyTorch)

ui-test

Playwright

Opt-in browser tests of the explorer (OWRT_RUN_GUI_TESTS=1)

dev

pytest, ruff

Running the test suite

docs

Sphinx, pydata theme, MyST-NB

Building this site

Next step#

Continue with the Quickstart, or start the browser explorer as described in Web explorer.