Fuser

A Fusion Code Generator for NVIDIA GPUs (commonly known as "nvFuser")

PyPI Installation

PyPI: https://pypi.org/project/nvfuser

nvFuser provides pre-built wheels for Python 3.10 and 3.12, available through multiple channels depending on your PyTorch version requirements.

Nightly Builds

Nightly nvFuser wheels are built against PyTorch:nightly and published to https://pypi.nvidia.com:

pip install --pre nvfuser-cuXXY --extra-index-url https://pypi.nvidia.com

[!note] nvFuser supports CUDA 12.6+. cuXXY denotes the CUDA major XX and minor Y version. If you have CUDA 12.8 use nvfuser-cu128.

To install nvFuser with a compatible PyTorch nightly build:

pip install --pre "nvfuser-cu128[torch]" --extra-index-url https://pypi.nvidia.com

[!warning] Installing with the [torch] extra will replace your existing PyTorch installation with a compatible nightly build.

Stable Releases

Stable wheels are built against PyTorch stable releases and published to both https://pypi.org and https://pypi.nvidia.com. Select the package matching your CUDA Toolkit version:

pip install nvfuser-cu128-torch29

Releases are published on the 1st and 15th of each month, and when significant changes are introduced. For legacy versions, see PyPI.

Recommendation: Use the latest nvFuser build with the most recent CUDA Toolkit and PyTorch versions for optimal performance and features.

[!important] Stable nvFuser release wheels are not guaranteed to be compatible with PyTorch nightly builds. Select the appropriate package for your environment.

Building From Source

Required:

• C++20 compliant compiler:
  • GCC >= 13.1 or Clang >= 19
• Python >= 3.10
• CMake >= 3.18
• Ninja
• CUDA Toolkit >= 12.6 (recommend 12.8+)
• PyTorch >= 2.9 (recommend latest stable/nightly release)
• pybind11 >= 3.0
• LLVM >= 18.1

[!note]

• PyTorch MUST be built w/ CUDA support.
• The PyTorch CUDA version MUST match the CUDAToolkit version.

Optional:

• nvidia-matmul-heuristics (enhanced matmul scheduling)

Build Steps

1. Clone the repository and initialize submodules:

git clone --recursive https://github.com/NVIDIA/Fuser.git
cd Fuser

If you already cloned without --recursive, initialize submodules:

git submodule update --init --recursive

2. Install Python dependencies:

pip install -r requirements.txt

3. Build and install nvFuser:

pip install --no-build-isolation -e python -v

The build system will automatically validate all dependencies and provide helpful error messages if anything is missing.

Build Options

You can customize the build using environment variables:

Build Configuration:

• MAX_JOBS=<n> - Control compilation parallelism (e.g., MAX_JOBS=8)
• NVFUSER_BUILD_BUILD_TYPE - Build in (Debug/RelWithDebInfo/Release) mode.
• NVFUSER_BUILD_DIR=<path> - Custom build directory (default: ./python/build)
• NVFUSER_BUILD_INSTALL_DIR=<path> - Custom install directory (default: ./nvfuser)

Build Targets:

• NVFUSER_BUILD_NO_PYTHON=1 - Skip Python bindings.
• NVFUSER_BUILD_NO_TEST=1 - Skip C++ tests.
• NVFUSER_BUILD_NO_BENCHMARK=1 - Skip benchmarks.

Advanced Options:

• NVFUSER_BUILD_WITH_UCC=1 - Enable UCC support for multi-device operations.
• NVFUSER_BUILD_WITHOUT_DISTRIBUTED=1 - Build without multi-device support.
• NVFUSER_BUILD_CPP_STANDARD=<n> - Specify C++ standard (default: 20).

Example with custom options:

MAX_JOBS=8 NVFUSER_BUILD_BUILD_TYPE=Debug pip install --no-build-isolation -e python -v

Verifying the Installation

Test your installation with a simple fusion:

python -c "import nvfuser; print('nvFuser successfully imported from:', nvfuser.__file__)"

Run the Python test suite:

pytest tests/python/

Run C++ tests (if built):

./build/bin/test_nvfuser