PennyLane Keras Layer

A library to introduce Keras 3 support to PennyLane with full support for multi-backend training.

Overview

This package provides integration between PennyLane (a quantum computing framework) and Keras 3, enabling the use of quantum layers in neural networks with support for multiple backends.

Key Features

• 🔄 Multi-Backend Support: Works seamlessly with JAX, TensorFlow, and PyTorch
• 🎯 Easy Integration: Drop-in replacement for standard Keras layers
• ⚡ High Performance: JIT compilation support with JAX backend
• 💾 Serialization: Full support for model saving and loading
• 🔧 Flexible Configuration: Customizable quantum circuit parameters
• 📊 Data Re-Uploading: Implements powerful quantum ML paradigm

Installation

From source

git clone https://github.com/vinayak19th/pennylane-keras-layer.git
cd pennylane-keras-layer
pip install -e .

Development installation

For development, install with additional dependencies:

pip install -e ".[dev]"

Requirements

• Python >= 3.8
• PennyLane >= 0.30.0
• Keras >= 3.0.0
• NumPy >= 1.21.0

Quick Start

Option 1: Data Re-Uploading Circuit Layer

import os
os.environ["KERAS_BACKEND"] = "jax"  # Set backend before importing keras

import numpy as np
import keras
from pennylane_keras_layer import KerasDRCircuitLayer

# Create sample data
X_train = np.linspace(-np.pi, np.pi, 100).reshape(-1, 1)
y_train = np.sin(X_train)

# Create a Data Re-Uploading quantum layer
q_layer = KerasDRCircuitLayer(layers=3, scaling=1.0, num_wires=1)

# Build a model
model = keras.Sequential([
    keras.layers.Input(shape=(1,)),
    q_layer
])

# Compile and train
model.compile(optimizer="adam", loss="mse")
model.fit(X_train, y_train, epochs=50)

Option 2: Generic Circuit Layer with Custom QNode

import pennylane as qml
from pennylane_keras_layer import KerasCircuitLayer

# Define a custom quantum circuit
dev = qml.device('default.qubit', wires=2)

@qml.qnode(dev)
def custom_circuit(weights, inputs):
    qml.RX(inputs[0], wires=0)
    qml.Rot(*weights[0], wires=0)
    qml.CNOT(wires=[0, 1])
    qml.Rot(*weights[1], wires=1)
    return qml.expval(qml.PauliZ(0))

# Create a generic quantum layer
weight_shapes = {"weights": (2, 3)}
q_layer = KerasCircuitLayer(custom_circuit, weight_shapes, output_dim=1)

# Use in a Keras model
model = keras.Sequential([
    keras.layers.Input(shape=(1,)),
    q_layer
])

Documentation

Comprehensive documentation is available in the docs/ directory:

• Getting Started: Installation and setup guide
• User Guide: In-depth concepts and best practices
• API Reference: Detailed API documentation
• Tutorials: Step-by-step tutorials for different backends
• Examples: Complete working examples
• FAQ: Frequently asked questions
• Contributing: How to contribute to the project

Examples

Check out the examples/ directory for usage examples:

# Run basic example
python examples/basic_example.py

# Run Fourier series approximation demo
python examples/fourier_series_example.py

Development

Running tests

The package includes a comprehensive test suite in the tests/ directory:

# Run all tests
pytest tests/

# Run with coverage
pytest tests/ --cov=pennylane_keras_layer --cov-report=term-missing

# Run specific test file
pytest tests/test_layer.py -v

# Run only integration tests
pytest tests/ -m integration

# Run tests with specific backend
pytest tests/ --backend=tensorflow  # or torch, or jax

Testing with different backends:

• TensorFlow: pytest tests/ --backend=tensorflow (requires pip install tensorflow)
• PyTorch: pytest tests/ --backend=torch (requires pip install torch)
• JAX: pytest tests/ --backend=jax (requires pip install jax jaxlib)

See tests/README.md for detailed information about the test suite.

Code formatting

black pennylane_keras_layer/ tests/

Linting

flake8 pennylane_keras_layer/ tests/

Contributing

Contributions are welcome! Please see our Contributing Guide for:

• How to report bugs
• How to suggest features
• Development setup
• Code standards
• Pull request process

Feel free to submit a Pull Request or open an issue!

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• PennyLane - Quantum machine learning framework
• Keras - Deep learning API