Unified XAI Documentation

PyPI Version Python Versions Documentation Status Build Status

A comprehensive, production-ready library for explaining deep learning models across multiple frameworks and modalities.

🚀 Getting Started

New to Unified XAI? Start with our quickstart guide.

📚 API Reference

Complete API documentation for all modules.

🎓 Tutorials

Step-by-step guides for various use cases.

🔬 Methods

Detailed explanations of all XAI algorithms.

✨ Key Features

Framework Agnostic: Works with PyTorch, TensorFlow, Keras, ONNX.
Multiple Modalities: Images, text, tabular data, time series.
Rich Visualizations: Heatmaps, interactive plots, dashboards.
Comprehensive Metrics: Faithfulness, stability, complexity.
High Performance: GPU acceleration, caching, parallelization.
Production Ready: Type hints, testing, error handling, logging.

Installation

pip install unified-xai

pip install unified-xai[torch]

pip install unified-xai[tf]

pip install unified-xai[all]

git clone https://github.com/yourusername/unified-xai.git
cd unified-xai
pip install -e ".[dev]"

Quick Example

from unified_xai import XAIAnalyzer, XAIConfig
from unified_xai.config import Framework, Modality

# Load your model
model = load_model('path/to/model')

# Configure XAI
config = XAIConfig(
    framework=Framework.PYTORCH,
    modality=Modality.IMAGE
)

# Initialize analyzer
analyzer = XAIAnalyzer(model, config)

# Generate explanation
explanation = analyzer.explain(
    input_data,
    method='integrated_gradients'
)

# Visualize
fig = analyzer.visualize(explanation, input_data)

API Reference

See full API in XAIAnalyzer, XAIConfig, and explanation result objects.

Supported Methods

Gradient-based: Vanilla Gradient, Integrated Gradients, SmoothGrad, Grad-CAM
Perturbation-based: LIME, SHAP
Attention-based: Transformer attention rollout