layerlens 0.1.0

A library for layer-by-layer explainability of deep learning models

LayerLens

LayerLens is a powerful library for layer-by-layer explainability of deep learning models. It provides insights into how models make decisions by analyzing the behavior of individual layers.

Features

• Layer-wise Explanation: Generate surrogate models for each layer to explain its behavior
• Hierarchical Insights: Stitch together layer explanations to understand the full model
• Interactive Visualization: Explore model behavior through an intuitive dashboard
• Production Monitoring: Detect drift and localize failures in deployed models
• Framework Agnostic: Works with PyTorch, TensorFlow, and other major frameworks
• Comprehensive Testing: Full test suite ensuring reliability and stability

Installation

From Source (Recommended)

# Clone the repository
git clone https://github.com/your-username/LayerLens.git
cd LayerLens

# Install dependencies
pip install -r requirements.txt

# Install LayerLens
pip install -e .

Via pip (when published)

pip install layerlens

Note: You'll also need to install at least one ML framework:

pip install torch  # For PyTorch support
# or
pip install tensorflow  # For TensorFlow support

Quick Start

from layerlens import Explainer
import numpy as np

# Load your trained model (PyTorch or TensorFlow)
# model = ... (your trained model)

# Create a LayerLens explainer
explainer = Explainer(model)

# Prepare your data
input_data = np.random.rand(10, 28, 28, 1)  # Example

# Generate explanations
explanation = explainer.explain(input_data)

# Visualize the results
explainer.visualize(explanation, output_dir='./explanations')

# Monitor for drift (production use)
explainer.monitor(input_data)

Running Examples

Try the MNIST demo to see LayerLens in action:

cd examples
jupyter notebook mnist_demo.ipynb

Verification

Verify your installation is working correctly:

python tests/verify_installation.py

Run the full test suite:

pytest tests/

Documentation

For detailed documentation and usage guides:

• USAGE.md - Complete usage guide with examples
• docs/ - Technical documentation and theory
• examples/ - Jupyter notebook demos

Project Structure

LayerLens/
├── layerlens/           # Main package
│   ├── core/           # Core explainability engine
│   ├── visualization/  # Interactive dashboards and plots
│   ├── monitoring/     # Production monitoring tools
│   └── utils/          # Helper utilities
├── examples/           # Demo notebooks
├── tests/             # Test suite
└── docs/              # Documentation

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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

Citation

If you use LayerLens in your research, please cite:

@software{layerlens2024,
  title = {LayerLens: Layer-by-Layer Explainability for Deep Learning},
  author = {LayerLens Team},
  year = {2024},
  url = {https://github.com/your-username/LayerLens}
}

D:/GitHub/LayerLens/.venv/Scripts/python.exe -m jupyter notebook mnist_demo.ipynb