LightNumPy

Lightweight and Fast Numerical Computing Library for Python

A lightweight version of NumPy or Cupy (or similar functionality).

Install|Try LightNumPy - Doc: https://scikit-plots.github.io/dev/devel/index.html

User Installation:

Installing from source (REQUIRED OS/LIB BUILD PACKAGES)

You can also install lightnumpy from source if you want to take advantage of the latest changes: ```sh

Forked repo: https://github.com/scikit-plots/lightnumpy.git

git clone https://github.com/YOUR-USER-NAME/lightnumpy.git cd lightnumpy

bash docker/script/safe_dirs.sh # add safe directories for git git submodule update --init # download submodules

pip install -r ./requirements/all.txt

pip install -r ./requirements/build.txt pip install --no-build-isolation --no-cache-dir -e . -v ```

It is also possible to include optional dependencies:

```sh

(Optionally) Try Development [dev,build,test,docs,gpu]

gpu refer NVIDIA CUDA support like cupy lib

python -m pip install --no-build-isolation --no-cache-dir -e .[dev,build,test,docs] -v

gpu refer NVIDIA CUDA support like cupy lib

pip install "lightnumpy[gpu]"

```

Installing using pip

The easiest way to set up LightNumPy is to install it using pip with the following command:

• by pypi: sh pip install lightnumpy -U

• by GITHUB use <branches> or <tags> If any:

  • Branches: bash #pip install git+https://github.com/scikit-plots/scikit-plots.git@<branches> # Latest in Development with C,Cpp,Fortran Support pip install git+https://github.com/scikit-plots/lightnumpy.git@main pip install git+https://github.com/scikit-plots/lightnumpy.git@maintenance/0.0.x
    
  • Tags: bash #pip install git+https://github.com/scikit-plots/scikit-plots.git@<tags> pip install git+https://github.com/scikit-plots/lightnumpy.git@v0.0.x

Why LightNumPy?

• Performance-Driven: Optimized for both CPU and hardware accelerators (GPU/TPU).
• Focused: Includes essential features without unnecessary overhead.
• Adaptable: Modular structure for customized extensions.
• Scalable: Ideal for IoT, embedded systems, and resource-limited devices.

See Also

• [1] https://scikit-plots.github.io/stable/api/scikitplot.numcppapi.html

- [2] https://github.com/dpilger26/NumCpp

LightNumPy Project Structure

sh lightnumpy/ │ ├── lightnumpy/ # Core library source code │ ├── __init__.py # Main package initializer │ ├── .clang-format # Code formatting rules for C/C++, code formatting rules like braces placement, and spacing. │ ├── .clang-tidy # Code linting rules for C/C++, code analysis, warnings, and bug detection. │ ├── _core/ # Low-level C and Higher-level C++ implementation sources │ │ ├── include/ # C and C++ headers │ │ │ ├── lightnumpy # lightnumpy core C and C++ headers │ │ │ └── NumCpp # NumCpp Headers [2] immutable │ │ └── src/ # C and C++ source files │ │ ├── dummymodule.c # Sample Module │ │ ├── hello.c # Print and string operations │ │ └── nc_version.cpp # NumCpp Headers versions │ ├── _gpu_core/ # GPU operations │ │ ├── include/ # GPU headers │ │ │ ├── lightnumpy # lightnumpy gpu core C and C++ headers as .cuh │ │ │ └── numC++ # numC++ Headers │ │ └── src/ # GPU source files │ │ ├── gpu_ops.cu # CUDA implementations │ │ └── cuda_helpers.cu # CUDA utility functions │ ├── _tpu_core/ # TPU operations │ │ ├── include/ # TPU headers │ │ │ ├── tpu_ops.hpp # TPU operation definitions │ │ │ └── tpu_helpers.hpp # TPU helper utilities │ │ └── src/ # TPU source files │ │ ├── tpu_ops.cpp # TPU operation implementations (via XLA) │ │ └── tpu_helpers.cpp # TPU utility functions │ ├── cy_bindings/ # Cython implementation, Cython bridging native libraries with Python APIs │ │ ├── __init__.py # Initialize the cython package │ │ ├── include/ # Cython Headers │ │ └── src/ # Cython Source files │ │ └── cython_helpers.pxd # Shared Cython declarations (optional) │ ├── py_bindings/ # Bindings for Python and native code, pybind11 bridging native libraries with Python APIs │ │ ├── include/ # pybind11 Headers │ │ └── src/ # pybind11 Source files │ │ ├── py_math.cpp # Element-wise operations │ │ ├── py_nc_random # Random array functionality pybind11 with NumCpp │ │ └── pybind_utils.cpp # Helper functions for bindings │ ├── python_api/ # Pure Python layer providing user-friendly interfaces for core functionality │ │ ├── __init__.py # API entry point for `python_api` │ │ ├── _utils_impl.py # get_include for lightnumpy library's C and C++ _core Headers with NumCpp Headers [2] │ │ ├── array.py # Array class implementation and basic methods │ │ ├── core.py # Contains core array functionality │ │ ├── linalg.py # Basic linear algebra operations (e.g., dot, transpose) │ │ ├── operations.py # Element-wise operations (e.g., addition, multiplication) (CPU, GPU, TPU) │ │ ├── gpu_operations.py # GPU-specific Python operations │ │ ├── tpu_operations.py # TPU-specific Python operations │ │ └── utils.py # Utility functions for array manipulation │ └── tests/ # Core library tests │ ├── test_array.py # Test for array module │ ├── test_tensor.py # Test for tensor module │ ├── test_gpu_ops.py # Test for GPU operations │ ├── test_tpu_ops.py # Test for TPU operations │ ├── test_cython_array.py # Test for Cython array implementation │ └── test_cython_utils.py # Test for Cython utility functions │ ├── examples/ # Example usage and demos │ ├── array_example.py # Example for arrays │ ├── tensor_example.py # Example for tensors │ ├── gpu_example.py # Example for GPU operations │ └── tpu_example.py # Example for TPU operations │ ├── docs/ # Documentation │ ├── index.md # Documentation index │ ├── api/ # API reference │ │ ├── gpu_api.md # GPU API documentation │ │ └── tpu_api.md # TPU API documentation │ └── developer_guide.md # Developer setup and guide │ ├── .github/ # CI/CD configuration │ ├── issue_templates/ # GitHub issue templates │ └── workflows/ # GitHub Actions workflows │ └── ci.yml # Main CI pipeline configuration │ ├── meson.build # Meson build configuration ├── LICENSE # Project license ├── pyproject.toml # Python project configuration └── README.md # Project overview ├── setup.cfg # Optional Python packaging configuration

FAQs

Q: How does lightnumpy differ from numpy?

A: While numpy is a full-featured numerical library, lightnumpy is a lightweight version, designed to be faster and more efficient for specific use cases. It focuses on essential features, leaving out less commonly used ones.

Q: Is lightnumpy compatible with numpy?

A: For most common operations, yes! It provides a familiar API to help you transition smoothly.

Q: Can I use lightnumpy for GPU/TPU tasks?

A: Absolutely! lightnumpy has built-in support for hardware accelerators.