EigenTorch

EigenTorch is a header-only Eigen-compatible API layer backed by libtorch tensors. It is designed for projects that want Eigen-style C++ APIs while executing calculations through PyTorch (CPU, CUDA, or Apple MPS).

What Is Implemented

• Core matrix/vector operations (Eigen::Matrix, dense arithmetic, blocks, maps, arrays)
• Dense decompositions and solvers (LU, QR, SVD, eigenvalue methods, LLT/LDLT)
• Geometry stack (Quaternion, AngleAxis, Transform, Translation, Rotation2D, Scaling, ParametrizedLine, Hyperplane, AlignedBox)
• Sparse stack (SparseMatrix, SparseVector, SparseLU, SparseQR, SimplicialLLT, SimplicialLDLT)
• Expanded unsupported modules with libtorch-backed calculations:
  • AdolcForward
  • AlignedVector3
  • ArpackSupport
  • AutoDiff
  • BVH
  • EulerAngles
  • FFT
  • IterativeSolvers
  • KroneckerProduct
  • LevenbergMarquardt
  • MatrixFunctions
  • MPRealSupport
  • NNLS
  • NonLinearOptimization
  • NumericalDiff
  • OpenGLSupport
  • Polynomials
  • SparseExtra
  • SpecialFunctions
  • Splines
  • CXX11/Tensor
  • CXX11/TensorSymmetry
  • CXX11/ThreadPool

Build

From repository root:

cmake -S . -B build
cmake --build build -j8

Install

cmake --install build

Then from a consumer CMake project:

find_package(Torch REQUIRED)
find_package(Eigen REQUIRED)

target_link_libraries(my_app PRIVATE Eigen::Eigen)

Basic Usage

#include <Eigen/Dense>

int main() {
    Eigen::MatrixXf A = Eigen::MatrixXf::Random(3, 3);
    Eigen::VectorXf b = Eigen::VectorXf::Ones(3, 1);

    Eigen::VectorXf x = A.partialPivLu().solve(b);
    Eigen::MatrixXf C = A * A.inverse();
    (void)x;
    (void)C;
    return 0;
}

Device Selection

Set the global default device before constructing matrices:

#include <Eigen/Core>

int main() {
    if (torch::cuda::is_available()) {
        Eigen::DeviceManager::set_default_device(torch::kCUDA);
    } else if (Eigen::DeviceManager::is_mps_available()) {
        Eigen::DeviceManager::set_default_device(torch::kMPS);
    } else {
        Eigen::DeviceManager::set_default_device(torch::kCPU);
    }
    return 0;
}

Test Suite

Run all tests:

cd build
ctest --output-on-failure

Current test targets:

• core_test
• sparse_linalg_test
• api_compat_test
• unsupported_compat_test
• unsupported_full_modules_test

Documentation

Sphinx sources are in docs/.

Generate docs:

cd docs
doxygen Doxyfile
make html

Main page after build: docs/_build/html/index.html.