pyro
Deep universal probabilistic programming with Python and PyTorch
What is pyro?
Deep universal probabilistic programming with Python and PyTorch
Overview of pyro
What it is
Pyro is a flexible, scalable deep probabilistic programming library built on PyTorch. It was originally developed at Uber AI and is now actively maintained by community contributors, including a dedicated team at the Broad Institute. In 2019, Pyro became a project of the Linux Foundation. It is designed to be a universal probabilistic programming language (PPL) that can represent any computable probability distribution, while scaling to large data sets with little overhead compared to hand-written code.
Key features
- Universal: Pyro can represent any computable probability distribution.
- Scalable: It scales to large data sets with minimal overhead.
- Minimal: Implemented with a small core of powerful, composable abstractions.
- Flexible: Provides high-level abstractions for generative and inference models, while allowing experts to customize inference.
Use cases
Pyro is suitable for probabilistic modeling and inference tasks, such as experimental design and time-to-event modeling. It can be used in any domain where deep probabilistic programming is applicable, including but not limited to Bayesian statistics, machine learning, and artificial intelligence research.
Getting started
To install a stable release via pip:pip install pyro-pplTo install from source:git clone [email protected]:pyro-ppl/pyro.gitcd pyrogit checkout masterpip install .To install with extra packages for examples/tutorials:pip install pyro-ppl[extras]For the dev branch, install via pip:pip install git+https://github.com/pyro-ppl/pyro.gitOr from source:git clone https://github.com/pyro-ppl/pyrocd pyropip install .
License and pricing
Pyro is licensed under the Apache-2.0 license and is open source.
Based on the project's public README and repository metadata.