aesara-devs/aesara

Health v2 · maintenance only

Flags: archived no_license

Adoption not part of the score

1219 stars · 152 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Aesara is a Python library for defining, optimizing, and efficiently evaluating mathematical expressions over multi-dimensional arrays, forked from Theano. It provides a hackable symbolic graph framework with automatic differentiation and transpilation to C, JAX, and Numba.

Use cases

• define and optimize symbolic tensor expressions in python
• compute gradients of mathematical expressions automatically
• compile numpy-like expressions to fast C, JAX, or Numba code
• build custom operators with graph rewrites and optimizations
• probabilistic programming backend for tensor computation

When to choose

• you need a hackable pure-Python symbolic tensor framework
• you want Theano-style graph optimization and autodiff
• you need to transpile symbolic graphs to multiple backends

When to avoid

• you need an actively maintained project - Aesara is no longer maintained
• you want deep-learning training at scale - use PyTorch or JAX instead
• starting a new project - consider its successor PyTensor

Facets

library · maturity abandoned

machine-learning compiler math serialization machine-learning performance data-science python cross-platform symbolic-computation automatic-differentiation tensor-library theano-fork term-rewriting transpilation numba jax algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="aesara-devs/aesara")