pylibann 0.1.0__tar.gz

pylibann-0.1.0/.github/FUNDING.yml

@@ -0,0 +1,3 @@
+# These are supported funding model platforms
+
+github: mseminatore

pylibann-0.1.0/.github/copilot-instructions.md

@@ -0,0 +1,48 @@
+# PyANN - Copilot Instructions
+
+## Architecture
+
+PyANN is a Python wrapper around the [libann](https://github.com/mseminatore/ann) C neural network library, exposing a Keras-like API. It has two distinct layers:
+
+- **C library** (`ann/` submodule) — core neural network engine built as a shared library (`.dll`/`.so`/`.dylib`). Bindings are declared in `src/pyann/_bindings/` using CFFI. The C library has its own guidelines in `ann/.github/copilot-instructions.md`.
+- **Python package** (`src/pyann/`) — high-level API wrapping the C library. Key modules:
+  - `core/network.py` — `Sequential` model (the main entry point)
+  - `core/layers.py` — `Dense`, `Input` layers
+  - `core/tensor.py` — Tensor wrapper around C tensors
+  - `callbacks.py` — LR schedulers (`StepDecay`, `ExponentialDecay`, `CosineAnnealing`) and `EarlyStopping`
+  - `hypertune/` — hyperparameter search (`GridSearch`, `RandomSearch`, `BayesianSearch`, `TPESearch`)
+  - `utils/` — data loading, normalization, one-hot encoding, NumPy compatibility (`compat.py`)
+
+NumPy is optional. The `utils/compat.py` module provides `has_numpy`, `to_array`, and `get_shape` helpers that work with plain Python lists as a fallback.
+
+## Build & Test
+
+```bash
+# Build the C shared library (requires CMake)
+python build_lib.py              # Release build
+python build_lib.py --cblas      # With CBLAS acceleration
+python build_lib.py --clean      # Clean rebuild
+
+# Install Python package in dev mode
+pip install -e ".[dev]"
+
+# Run all tests
+pytest
+
+# Run a single test file
+pytest tests/test_layers.py
+
+# Run a single test
+pytest tests/test_layers.py::TestDense::test_basic_dense -v
+```
+
+The built shared library is copied to `src/pyann/lib/`. You can also set `PYANN_LIB_PATH` to point to a custom library location.
+
+## Conventions
+
+- **src layout**: Python package lives under `src/pyann/`, not at the project root.
+- **Enum + string API**: Activations, losses, and optimizers accept both enum values (`Activation.RELU`) and string names (`"relu"`). Parsing helpers (`parse_optimizer`, `parse_loss`) handle conversion.
+- **Error handling**: C library errors are translated via `raise_for_error_code()` in `exceptions.py`. Python-side validation uses standard `ValueError`/`TypeError`.
+- **Test style**: Tests use pytest with class-based grouping (`class TestDense:`). Each test class covers one component.
+- **Type hints**: All public APIs use type hints. `TYPE_CHECKING` guard is used for numpy imports.
+- **C function prefixes**: `ann_*` for network functions, `tensor_*` for tensor operations (relevant when modifying `_bindings/` CFFI definitions).

pylibann-0.1.0/.github/workflows/blas-bench.yml

@@ -0,0 +1,96 @@
+name: BLAS benchmark
+
+# Compares libann built with no BLAS, the project's CBLAS, OpenBLAS (and Accelerate on macOS)
+# Manual only: a paths filter on pull_request matches the whole PR diff, so it
+# would run on every push to a PR that ever touched benchmarks/
+on:
+  workflow_dispatch:
+
+jobs:
+  bench:
+    name: ${{ matrix.name }}
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - { name: macos-arm64,   os: macos-14 }
+          - { name: linux-x86_64,  os: ubuntu-latest }
+          - { name: linux-aarch64, os: ubuntu-24.04-arm }
+    env:
+      OUT: ${{ github.workspace }}/bench-libs
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install OpenBLAS (Linux)
+        if: runner.os == 'Linux'
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libopenblas-dev
+          # libann looks for include/cblas.h and lib/libopenblas under one root.
+          # libopenblas-dev is a meta-package, so locate the real files with find.
+          hdr=$(find /usr/include -name cblas.h -path '*openblas*' | head -1)
+          lib=$(find /usr/lib -name 'libopenblas.so' | head -1)
+          echo "header: $hdr  library: $lib"
+          sudo mkdir -p /opt/OpenBLAS/lib
+          sudo ln -sfn "$(dirname "$hdr")" /opt/OpenBLAS/include
+          sudo ln -sf "$lib" /opt/OpenBLAS/lib/libopenblas.so
+          ls -lL /opt/OpenBLAS/include/cblas.h /opt/OpenBLAS/lib/libopenblas.so
+
+      - name: Install OpenBLAS (macOS)
+        if: runner.os == 'macOS'
+        run: |
+          brew install openblas
+          sudo mkdir -p /opt
+          sudo ln -sfn "$(brew --prefix openblas)" /opt/OpenBLAS
+
+      - name: Build and install CBLAS from main
+        run: |
+          git clone --depth 1 https://github.com/mseminatore/cblas /tmp/cblas
+          echo "CBLAS commit: $(git -C /tmp/cblas rev-parse HEAD)"
+          # PIC is required to link the static library into libann's shared library
+          cmake -S /tmp/cblas -B /tmp/cblas/build -DCMAKE_BUILD_TYPE=Release \
+                -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DCMAKE_INSTALL_PREFIX=/opt/cblas
+          cmake --build /tmp/cblas/build --target cblas --config Release
+          sudo cmake --install /tmp/cblas/build --config Release
+
+      - name: Build libann variants
+        run: |
+          build() {  # name, cmake flags...
+            name=$1; shift
+            cmake -S . -B build-$name -DCMAKE_BUILD_TYPE=Release "$@"
+            cmake --build build-$name --config Release
+            mkdir -p $OUT/$name
+            cp -L src/pyann/lib/libann.* $OUT/$name/
+          }
+          build none
+          build cblas -DUSE_CBLAS=ON
+          build openblas -DUSE_BLAS=ON
+          if [ "$RUNNER_OS" = macOS ]; then build accelerate -DUSE_ACCELERATE=ON; fi
+
+      - name: Run benchmarks
+        run: |
+          python -m pip install numpy cffi
+          # A crash must be visible: no pipes, per-variant exit codes, faulthandler
+          for dir in $OUT/*/; do
+            variant=$(basename $dir)
+            echo "::group::$variant"
+            PYANN_LIB_PATH=$dir PYTHONPATH=src python -X faulthandler benchmarks/bench_blas.py \
+              --variant $variant --out results.jsonl || echo "::error::$variant exited with status $?"
+            echo "::endgroup::"
+          done
+          {
+            echo "## ${{ matrix.name }} ($(python -c 'import os; print(os.cpu_count())') CPUs)"
+            python benchmarks/summarize.py < results.jsonl
+          } | tee -a "$GITHUB_STEP_SUMMARY"
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: bench-${{ matrix.name }}
+          path: results.jsonl

pylibann-0.1.0/.github/workflows/wheels.yml

@@ -0,0 +1,128 @@
+name: Build wheels
+
+on:
+  push:
+    branches: [main]
+    tags: ["v*"]
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  sdist:
+    name: Source distribution
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Build sdist
+        run: |
+          python -m pip install build
+          python -m build --sdist
+
+      - name: Check package metadata
+        run: |
+          python -m pip install twine
+          twine check dist/*
+
+      # Prove the sdist is self-contained: build and test it away from the checkout
+      - name: Install and test sdist
+        env:
+          PYANN_EXPECT_BLAS: openblas
+        run: |
+          python -m pip install pytest numpy dist/*.tar.gz
+          mkdir /tmp/sdist-test && cp -r tests /tmp/sdist-test/
+          cd /tmp/sdist-test && python -m pytest tests
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: sdist
+          path: dist/*.tar.gz
+
+  wheels:
+    name: Wheel (${{ matrix.name }})
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - { name: macos-arm64,    os: macos-14,           archs: arm64 }
+          - { name: linux-x86_64,   os: ubuntu-latest,      archs: x86_64 }
+          - { name: linux-aarch64,  os: ubuntu-24.04-arm,   archs: aarch64 }
+          - { name: windows-x86_64, os: windows-latest,     archs: AMD64 }
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - uses: pypa/cibuildwheel@v2.23.3
+        env:
+          CIBW_ARCHS: ${{ matrix.archs }}
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: wheel-${{ matrix.name }}
+          path: wheelhouse/*.whl
+
+  # Releases are tag-driven: push vX.Y.Z to publish to PyPI, or a pre-release
+  # tag such as vX.Y.ZrcN to publish to TestPyPI.
+  release-meta:
+    name: Release metadata
+    if: startsWith(github.ref, 'refs/tags/v')
+    runs-on: ubuntu-latest
+    outputs:
+      prerelease: ${{ steps.meta.outputs.prerelease }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Check tag against package version
+        id: meta
+        run: |
+          python -m pip install packaging
+          python - <<'PY' >> "$GITHUB_OUTPUT"
+          import os, tomllib
+          from packaging.version import Version
+
+          tag = os.environ["GITHUB_REF_NAME"].removeprefix("v")
+          with open("pyproject.toml", "rb") as f:
+              version = tomllib.load(f)["project"]["version"]
+          if Version(tag) != Version(version):
+              raise SystemExit(f"::error::tag v{tag} does not match pyproject.toml version {version}")
+          print(f"prerelease={str(Version(version).is_prerelease).lower()}")
+          PY
+
+  publish:
+    name: Publish
+    needs: [sdist, wheels, release-meta]
+    runs-on: ubuntu-latest
+    # Each index has its own trusted publisher bound to one of these environments
+    environment: ${{ needs.release-meta.outputs.prerelease == 'true' && 'testpypi' || 'pypi' }}
+    permissions:
+      id-token: write   # trusted publishing: no API token needed
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          path: dist
+          merge-multiple: true
+
+      - name: List distributions
+        run: ls -l dist
+
+      - name: Publish to TestPyPI
+        if: needs.release-meta.outputs.prerelease == 'true'
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/
+
+      - name: Publish to PyPI
+        if: needs.release-meta.outputs.prerelease != 'true'
+        uses: pypa/gh-action-pypi-publish@release/v1

pylibann-0.1.0/.gitignore

@@ -0,0 +1,217 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# data files
+fashion-mnist_train.csv
+fashion-mnist_test.csv
+fashion_mnist.nna
+
+# C extensions
+*.so
+*.dll
+*.dylib
+
+# Built shared libraries
+src/pyann/lib/
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/

pylibann-0.1.0/.gitmodules

@@ -0,0 +1,3 @@
+[submodule "ann"]
+	path = ann
+	url = https://github.com/mseminatore/ann.git

pylibann-0.1.0/CLAUDE.md

@@ -0,0 +1,61 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+PyANN is a Python wrapper around the [libann](https://github.com/mseminatore/ann) C neural network library, exposing a Keras-like API. It has two layers:
+
+- **C library** (`ann/` git submodule) — core neural network engine built as a shared library via CMake. CFFI bindings are declared in `src/pyann/_bindings/`.
+- **Python package** (`src/pyann/`, src layout) — high-level API wrapping the C library. `Sequential` is the main entry point.
+
+NumPy is optional — `utils/compat.py` provides fallback helpers that work with plain Python lists.
+
+## Build & Test
+
+```bash
+# Build the native library and install in dev mode (scikit-build-core runs CMake)
+pip install -e ".[dev]"
+
+# Pass CMake options through CMAKE_ARGS, e.g. CUDA or a plain CPU build
+CMAKE_ARGS="-DUSE_CUDA=ON" pip install -e .
+CMAKE_ARGS="-DUSE_METAL=OFF -DUSE_ACCELERATE=OFF" pip install -e .
+
+# Or build only the shared library into src/pyann/lib/ (no install)
+python build_lib.py                  # defaults: no BLAS (see flags below)
+python build_lib.py --accelerate --metal   # macOS
+python build_lib.py --scipy-openblas       # Linux/Windows (pip install scipy-openblas32 first)
+python build_lib.py --cblas | --blas | --cuda | --clean
+
+# Run tests
+pytest                                                    # all tests
+pytest tests/test_layers.py                               # single file
+pytest tests/test_layers.py::TestDense::test_basic_dense -v  # single test
+
+# Build a wheel / sdist
+pip wheel . --no-deps -w dist
+python -m build --sdist
+```
+
+The `ann/` submodule must be checked out (`git submodule update --init`). Wheels default to Accelerate+Metal on macOS and the bundled `scipy-openblas32` OpenBLAS on Linux/Windows (see `[[tool.scikit-build.overrides]]` in `pyproject.toml`). CI (`.github/workflows/wheels.yml`) builds and tests wheels with cibuildwheel; `tests/test_backend.py` fails a wheel that lost its BLAS. `.github/workflows/blas-bench.yml` (manual) compares BLAS libraries using `benchmarks/`.
+
+The built shared library goes to `src/pyann/lib/` for dev builds. Override with `PYANN_LIB_PATH` env var.
+
+## Releasing
+
+The PyPI distribution is `pylibann` (the name `pyann` is taken by an unrelated project); the import name is still `pyann`. Releases are tag-driven via `.github/workflows/wheels.yml`:
+
+1. Set `version` in `pyproject.toml` (PEP 440) and merge to `main`.
+2. Push a tag `vX.Y.Z`. The tag must equal the version or the `release-meta` job fails.
+3. A pre-release tag (`vX.Y.ZrcN`) publishes to TestPyPI via the `testpypi` environment; a final tag publishes to PyPI via the `pypi` environment. Both use trusted publishing (no tokens); each index needs a publisher pointing at this repo, `wheels.yml`, and the matching environment name.
+
+Try a release candidate on TestPyPI first and install it in a clean venv before tagging the real release.
+
+## Conventions
+
+- **Enum + string API**: Activations, losses, and optimizers accept both enum values (`Activation.RELU`) and strings (`"relu"`). Parsing helpers (`parse_optimizer`, `parse_loss`, `parse_activation`) handle conversion.
+- **Error handling**: C errors are translated via `raise_for_error_code()` in `exceptions.py`. Python-side validation uses `ValueError`/`TypeError`.
+- **Test style**: pytest with class-based grouping (`class TestDense:`). Each class covers one component.
+- **Type hints**: All public APIs use type hints. `TYPE_CHECKING` guard for numpy imports.
+- **C binding prefixes**: `ann_*` for network functions, `tensor_*` for tensor operations (in `_bindings/` CFFI definitions).
+- **Python enum values** (`IntEnum`) must match C enum ordinals exactly.