Home

Info

illia is actively developed and maintained. While we're working toward full stability across all backends, the core functionality is production-ready. Join our community to stay updated!

Introduction

illia is a cutting-edge library for Bayesian Neural Networks that brings uncertainty quantification to deep learning. Designed with flexibility in mind, it seamlessly integrates with multiple backends and popular frameworks.

For full documentation, please visit the site: https://ericssonresearch.github.io/illia/

Why Choose illia?

• Multi-Backend Support: Works with PyTorch, TensorFlow, and JAX.
• Graph Neural Networks: Integrated with PyTorch Geometric, DGL, and Spektral.
• Developer Friendly: Intuitive API design and comprehensive documentation.

Quick Start

Get started with illia in just a few lines of code:

import os
import torch

# Configure backend (PyTorch is default)
os.environ["ILLIA_BACKEND"] = "torch"

import illia
from illia.nn import Conv2d

# Create a Bayesian convolutional layer
conv_layer = Conv2d(
    input_channels=3,
    output_channels=64,
    kernel_size=3,
    bias=True
)

# Forward pass with uncertainty
input_tensor = torch.rand(1, 3, 32, 32)
output_mean, output_std = conv_layer(input_tensor)

print(f"Output shape: {output_mean.shape}")
print(f"Uncertainty quantified: {output_std.mean():.4f}")

Contributing

We welcome contributions from the community! Whether you're fixing bugs, adding features, or improving documentation:

1. Read our contributing guide for development setup.
2. Check open issues for ways to help.
3. Submit bug reports using our issue templates.

License

illia is released under the MIT License. We hope you find it useful and inspiring for your projects!