KernelFoundry is a Python package for defining and evaluating GPU kernel tasks.
This repository provides the task-side test harness used in kernel generation workflows:
- A base task interface (
kernelfoundry.custom_test.CustomTest) for implementing task-specific build and pytest test logic. - Build helpers for compiling candidate kernels into PyTorch extensions (via Torch or
icpx). - Pytest fixtures for correctness/performance runs and collecting runtime data.
- Validation helpers (
assert_allclose, cosine similarity, and related utilities). - Runtime and machine-info helpers for benchmarking and metadata capture.
In other words, this package is what you use to author tasks and evaluate generated kernels. The external orchestration layers (for example, UI/services that generate candidates) are separate from this repo.
- Create a task test class by deriving from
kernelfoundry.custom_test.CustomTest. - Implement build logic (optional) and correctness/performance tests.
- Compile candidate kernel code with
compile_torch_extension(...). - Run pytest to validate correctness and collect benchmark timings.
KernelFoundry supports SYCL, OpenCL and CUDA kernels.
Install the package and basic dependencies with
pip install .
The package requires Python 3.10+ (we commonly use Python 3.12). Install into a virtual environment if desired.
The kernelfoundry package has minimal core dependencies. Depending on your task, you may also need framework/toolchain dependencies (for example PyTorch and GPU toolchains) to compile and test kernels locally.
To run PyTorch-based tasks, install torch appropriate for your hardware:
# on Intel hardware:
pip install torch==2.9.0 --index-url https://download.pytorch.org/whl/xpu
# on NVIDIA hardware:
pip install torch==2.9.0 --index-url https://download.pytorch.org/whl/cu129
API documentation is available and can be built using Sphinx. To build the documentation:
pip install .[docs]
cd docs
make html
The generated documentation will be in docs/_build/html/. See the docs/README.md for more information.