---
id: tfp-nightly
version: "0.26.0.dev20260814"
license: Apache 2.0
license_treatment: permissive
maintenance: active
---

# tfp-nightly — Probabilistic modeling and statistical inference in TensorFlow

License: permissive · Maintenance: active · Downloads: 274.9K/mo

## What it is and what it does

TensorFlow Probability is a library for building and reasoning about probabilistic models within the TensorFlow ecosystem. It layers statistical building blocks—distributions, bijectors, and transformations—on top of TensorFlow's numerical operations, then adds higher-level tools for model construction (joint distributions, probabilistic layers) and inference (MCMC, variational inference, stochastic optimizers). The library also works as a JAX substrate, letting you use the same probabilistic API with JAX's functional paradigm instead of TensorFlow's graph execution.

Typical use involves defining a probabilistic model using distributions and layers, then performing inference via gradient-based optimization or sampling-based methods. It is designed for researchers and practitioners building Bayesian neural networks, hierarchical models, and other probabilistic systems that benefit from automatic differentiation and hardware acceleration. The package depends on absl-py, six, numpy, decorator, cloudpickle, gast, and dm-tree.

Use it for:

- Build Bayesian neural networks with weight uncertainty and quantify prediction confidence
- Perform hierarchical statistical modeling and inference on grouped or nested data
- Implement Markov chain Monte Carlo or variational inference for posterior approximation
- Define and sample from complex joint probability distributions with dependencies
- Combine probabilistic layers with standard deep learning models for uncertainty-aware inference
- Conduct dimensionality reduction or clustering using probabilistic generative models

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

TensorFlow Probability provides probabilistic modeling, statistical inference, and Bayesian machine learning tools integrated with TensorFlow and JAX, including distributions, bijectors, inference algorithms, and probabilistic neural network layers.

Yes, if you are building probabilistic or Bayesian machine learning models in TensorFlow or JAX. The library is actively maintained, has no known vulnerabilities, and integrates with TensorFlow's ecosystem. Install with caution only if your project requires stable APIs, as the package is marked Beta and interfaces may change; verify compatibility with your TensorFlow version for production use.

## Install

pip install tfp-nightly
uv add tfp-nightly
poetry add tfp-nightly

## Installing tfp-nightly

Before you install: Low install friction with a pure-Python wheel and common dependencies. The package is actively maintained with recent commits and no known vulnerabilities, though it is marked Beta and interfaces may change.

License in practice: Licensed under Apache 2.0 (permissive), allowing commercial use, modification, and distribution with minimal restrictions beyond attribution and liability disclaimers.

Quickstart:

pip install tfp-nightly

import tensorflow_probability as tfp
dist = tfp.distributions.Normal(loc=0., scale=1.)
samples = dist.sample()

Requires TensorFlow or JAX installed separately; TensorFlow Probability is a library layer, not a standalone numerical backend.

Verify before relying:

- Whether JAX substrate support requires separate JAX installation or is bundled
- Performance characteristics and scalability limits for large-scale probabilistic models
- Compatibility with specific TensorFlow versions beyond Python 3.9, 3.10, 3.11, 3.12 support

## Package facts

- License: Apache 2.0 (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 274.9K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

probabilistic machine learning tensorflow, bayesian inference deep learning, markov chain monte carlo sampling, variational inference optimization, probabilistic neural networks, statistical distributions tensorflow, uncertainty quantification, bayesian-inference, probabilistic-modeling, uncertainty-quantification

[View on SkillFed](https://skillfed.io/packages/tfp-nightly) · [View on PyPI](https://pypi.org/project/tfp-nightly/)