Top AI Repos — open-source AI, indexed and scored

Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.

Star history, last 90 days
Date	Stars
2026-07-24	711
2026-07-25	711
2026-07-28	711
2026-07-30	711
2026-08-06	711

Star history, last 90 days

Date

Stars

2026-07-24

711

2026-07-25

711

2026-07-28

711

2026-07-30

711

2026-08-06

711

# Edward2 Edward2 is a _simple_ probabilistic programming language. It provides core utilities in deep learning ecosystems so that one can write models as probabilistic programs and manipulate a model's computation for flexible training and inference. It's organized as follows: * [`edward2/`](https://github.com/google/edward2/blob/main/edward2/): Library code. * [`examples/`](https://github.com/google/edward2/blob/main/examples): Examples. * [`experimental/`](https://github.com/google/edward2/blob/main/experimental): Active research projects. Are you upgrading from Edward? Check out the guide [`Upgrading_from_Edward_to_Edward2.md`](https://github.com/google/edward2/blob/main/Upgrading_From_Edward_To_Edward2.md). The core utilities are fairly low-level: if you'd like a high-level module for uncertainty modeling, check out the guide for [Bayesian Layers](https://github.com/google/edward2/tree/main/edward2/tensorflow/layers). We recommend the [Uncertainty Baselines](https://github.com/google/uncertainty-baselines) if you'd like to build on research-ready code. ## Installation We recommend the latest development version. To install, run ```sh pip install "edward2 @ git+https://github.com/google/edward2.git" ``` You can also install the latest stable version using the following. As a caveat, however, we very rarely update the stable version (this is a passion project maintained by part-timers and scheduling releases every so often sucks up time). ```sh pip install edward2 ``` Edward2 supports three backends: TensorFlow (the default), JAX, and NumPy ([see below to activate](#using-the-jax-or-numpy-backend)). Installing `edward2` does not automatically install any backend. To get these dependencies, use for example `pip install edward2[tensorflow]"`, replacing `tensorflow` for the appropriate backend. Sometimes Edward2 uses the latest changes from TensorFlow in which you'll need TensorFlow's nightly package: use `pip install edward2[tf- nightly]`. ## 1. Models as Probabilistic Programs ### Random Variables In Edward2, we use [`RandomVariables`](https://github.com/google/edward2/blob/main/edward2/tensorflow/random_variable.py) to specify a probabilistic model's structure. A random variable `rv` carries a probability distribution (`rv.distribution`), which is a TensorFlow Distribution instance governing the random variable's methods such as `log_prob` and `sample`. Random variables are formed like TensorFlow Distributions. ```python import edward2 as ed normal_rv = ed.Normal(loc=0., scale=1.) ## <ed.RandomVariable 'Normal/' shape=() dtype=float32 numpy=0.0024812892> normal_rv.distribution.log_prob(1.231) ## <tf.Tensor: id=11, shape=(), dtype=float32, numpy=-1.6766189> dirichlet_rv = ed.Dirichlet(concentration=tf.ones([2, 3])) ## <ed.RandomVariable 'Dirichlet/' shape=(2, 3) dtype=float32 numpy= array([[0.15864784, 0.01217205, 0.82918006], [0.23385087, 0.69622266, 0.06992647]], dtype=float32)> ``` By default, instantiating a random variable `rv` creates a sampling op to form the tensor `rv.value ~ rv.distribution.sample()`. The default number of samples (controllable via the `sample_shape` argument to `rv`) is one, and if the optional `value` argument is provided, no sampling op is created. Random variables can interoperate with TensorFlow ops: the TF ops operate on the sample. ```python x = ed.Normal(loc=tf.zeros(2), scale=tf.ones(2)) y = 5. x + y, x / y ## (<tf.Tensor: id=109, shape=(2,), dtype=float32, numpy=array([3.9076924, 4.588356 ], dtype=float32)>, ## <tf.Tensor: id=111, shape=(2,), dtype=float32, numpy=array([-0.21846154, -0.08232877], dtype=float32)>) tf.tanh(x * y) ## <tf.Tensor: id=114, shape=(2,), dtype=float32, numpy=array([-0.99996394, -0.9679181 ], dtype=float32)> x[1] # 2nd normal rv ## <ed.RandomVariable 'Normal/' shape=() dtype=float32 numpy=-0.41164386> ``` ### Probabilistic Models Probabilistic models in Edward2 are expressed as Python functions that instantiate one or more `Random