SIByLEngine 0.0.4

SIByLEngine is a toy renderer framework based on Vulkan, designed for rapid prototyping of real-time and offline graphics algorithms requiring GPU acceleration. It supports interoperation with Python and PyTorch, enabling potential of rapid development, differentiable computing, and integration with deep learning components.

Requirements

Our framework currently supports only Windows and Linux with an NVIDIA GPU.
To run some of our examples, the GPU must also support hardware ray tracing.

Install precompiled binaries

For Windows users, follow the steps below to install our framework. Please use the specified Python and Torch versions.
If you require compatibility with other Python and Torch versions, please build from the source code.

# create a new conda environment
conda create -n sibyl python=3.8.18
conda activate sibyl

# install pytorch version 2.4.0 with CUDA 11.8
pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu118

# install sibylengine
pip install sibylengine==0.0.4

# install additional packages for examples
pip install numpy==1.26.3

Meanwhile, we highly recommend installing the Vulkan SDK, as it helps with debugging.

Our examples assume that the Vulkan SDK is properly installed. If you prefer to work without Vulkan SDK, you need to comment out the line containing core.rhi.EnumContextExtension.DEBUG_UTILS in our common.py file.

Build from source

1. Clone this repository to a local folder, then open the folder in Visual Studio or VS Code.

2. Set up the Python environment:

   • Ensure it has numpy, torch, and nanobind installed.
   • You can create one using conda:

     # create an env called "myenv"
     conda create -n myenv python=3.9
     conda activate myenv
     # install numpy and nanobind
     pip install nanobind numpy
     # install pytorch version 2.4.0 with CUDA 11.8
     pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu118

   • Update the "Python_EXECUTABLE" variable in CMakePresets.json to point to the Python executable in this environment.

3. Configure and build the project:

   • Select your preferred configure preset and build the project.

4. Install the built package:

   • After compilation, a .pyd (Windows) or .so (Linux) file should be created under release\sibylengine, so:

     # Navigate to the `/release` folder.
     cd release
     # Activate the target Python environment.
     conda activate myenv
     # install the package
     pip install .