Uber Unveils Pyro: An Open Source Probabilistic Programming Language

Uber has always been about making it’s existing products better and bringing reliable transportation service to users.

But that’s not all they are about.

Uber is also investing efforts in autonomous cars and artificial intelligence. And why not? that’s the new fad, innit?

However, to give something back to the developer community and improve their products, Uber has announced the open source release of Pyro.

Pyro is Uber’s homegrown probabilistic programming language. The company says that it’s “a tool for deep probabilistic modeling, unifying the best of modern deep learning and Bayesian modeling.”

By open sourcing Pyro programming language, Uber aims to accelerate research and applications of different deep learning-related techniques and decentralize them. Currently, in alpha stage, it’s developed by Uber AI Labs. Stanford University is also using Pyro.

Written in Python, this language supports PyTorch in the backend. The key principles that form the basis of Pyro’s design are as follows:

• Universal: Represent any computable probability distribution.
• Scalable:  Scales to large data sets with little overhead.
• Minimal: Implemented with a small core of powerful, composable abstractions.
• Flexible: Aims for automation when you want it, control when you need it.

How to install Pyro programming language?

Pyro supports Python 2.7 and Python 3+. To install it via pip, run:

In Python 3, run:

Install from source:

In upcoming months, Pyro programming language is expected to change rapidly as a bigger part of the developer community participate in its development. Uber’s own priorities in technical development include additional techniques for estimating expectation of gradients, improving the abstractions for modeling and advanced usage, etc.

You can find Pyro’s open source code on its GitHub page. Do give it a try and leave your valuable feedback.