# tum-pbs/PhiFlow

A differentiable PDE solving framework for machine learning

Repository: https://github.com/tum-pbs/PhiFlow
Canonical: https://ross.abutalabs.com/products/phiflow
Language: Python
License: MIT
License Family: permissive
Topics: differentiable-simulations, fluid-simulations, deep-learning, neural-networks, pde-solver
Last push: 2026-07-16T20:11:37+00:00

## Health v2 (maintenance only)
Score: 72/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 92, release rhythm 29, longevity 100
- inputs: {"age_days": 2464, "days_push": 48, "days_rel": 396, "gap_med": 71, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1929, forks 233 (observed 2026-08-28T04:05:55.879548+00:00)

## What it is
PhiFlow is an open-source Python simulation toolkit for solving partial differential equations with support for optimization and machine learning workflows. It integrates with NumPy, PyTorch, JAX, and TensorFlow to provide end-to-end differentiable physics simulations, such as fluid dynamics, combined with neural network models.

## Use cases
- run differentiable fluid simulations for machine learning
- train neural networks with physics simulation in the loss loop
- solve PDEs with automatic differentiation
- build physics-based deep learning models
- simulate smoke and fluid dynamics in Python
- optimize simulation parameters via gradients

## When to choose
- you need gradients flowing through physics simulations into ML models
- you want a PDE solver that works with PyTorch, JAX, or TensorFlow
- you are doing research in physics-based deep learning or differentiable simulation

## When to avoid
- you need a general-purpose CFD tool with production-grade solver accuracy
- you only need non-differentiable simulation without ML integration
- you need simulation domains outside PDE-based physics like rigid body dynamics

## Facets
- artifact type: library
- maturity: active
- function: simulation, machine-learning, deep-learning, math
- domain: machine-learning, simulation, deep-learning
- platform: python, cross-platform
- tags: differentiable-simulation, pde-solver, fluid-simulation, physics-based-deep-learning, automatic-differentiation, scientific-computing, algorithms

## Member repositories
- tum-pbs/PhiFlow (main) score 72

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:55.879548+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:09:03.742956+00:00, confidence not recorded.
  - readme: https://github.com/tum-pbs/PhiFlow (fetched 2026-08-28T04:05:55.879548+00:00, sha 2a22319022ee)
  - registry_pypi: https://pypi.org/pypi/phiflow/json (fetched 2026-08-29T10:48:25.162971+00:00, sha b483bac1ae1b)
- Data as of 2026-08-30T08:39:29.467469+00:00.