TensorFlow Probability is a library for probabilistic reasoning and statistical analysis in TensorFlow. As part of the TensorFlow ecosystem, TensorFlow Probability provides integration of probabilistic methods with deep networks, gradient-based inference via automatic differentiation, and scalability to large datasets and models via hardware acceleration (e.g., GPUs) and distributed computation.

The library consists of the following modules:

• Distributions (tfp.distributions, tfp.trainable_distributions): Probability distributions with efficient, composable manipulations.
• Edward2 (tfp.edward2): A probabilistic programming language, which enables flexible probabilistic models and flexible computation for their training and testing.
• Layers (tfp.layers): Neural network layers with uncertainty over the functions they represent, extending TensorFlow Layers.
• Monte Carlo (tfp.mcmc, tfp.optimizers, tfp.monte_carlo): Algorithms for approximate Bayesian inference via sampling.
• Variational Inference (tfp.vi): Algorithms for approximate Bayesian inference via optimization.
• Examples (tfp.examples): End-to-end implementations of probabilistic reasoning using TensorFlow Probability.

TensorFlow Probability is under active development. Interfaces may change at any time.

To install the latest version, run the following:

pip install --user --upgrade tfp-nightly    # depends on tensorflow (CPU-only)

TensorFlow Probability depends on a current nightly release of TensorFlow (tf-nightly); the --upgrade flag ensures you'll automatically get the latest version.

We also provide a GPU-enabled package:

pip install --user --upgrade tfp-nightly-gpu  # depends on tensorflow-gpu (GPU enabled)

Currently, TensorFlow Probability does not contain any GPU-specific code. The primary difference between these packages is that tensorflow-probability-gpu depends on a GPU-enabled version of TensorFlow.

To force a Python 3-specific install, replace pip with pip3 in the above commands. For additional installation help, guidance installing prerequisites, and (optionally) setting up virtual environments, see the TensorFlow installation guide.

You can also install from source. This requires the Bazel build system.

# sudo apt-get install bazel git python-pip  # Ubuntu; others, see above links.
git clone https://github.com/tensorflow/probability.git
cd probability
bazel build --config=opt --copt=-O3 --copt=-march=native :pip_pkg
PKGDIR=$(mktemp -d)
./bazel-bin/pip_pkg $PKGDIR
pip install --user --upgrade $PKGDIR/*.whl

Access the library using

import tensorflow_probability as tfp

See the tfp.examples module for examples of end-to-end implementations. They can also be run under command line: for example, run

python -m tensorflow_probability.examples.vae

to train a variational auto-encoder to generate MNIST digits. See the examples/ directory for more details.

We're eager to collaborate with you! Feel free to open an issue on GitHub and/or send us your pull requests. See CONTRIBUTING.md for more details. This project adheres to TensorFlow's code of conduct. By participating, you are expected to uphold this code.