Package phi

Open-source simulation toolkit built for optimization and machine learning applications.

Use one of the following imports:

• from phi.flow import * for NumPy mode
• from phi.tf.flow import * for TensorFlow mode
• from phi.torch.flow import * for PyTorch mode
• from phi.jax.flow import * for Jax mode

Project homepage: https://github.com/tum-pbs/PhiFlow

Documentation overview: https://tum-pbs.github.io/PhiFlow

PyPI: https://pypi.org/project/phiflow/

Sub-modules

phi.field

The fields module provides a number of data structures and functions to represent continuous, spatially varying data …

phi.flow

Main PhiFlow import: from phi.flow import * …

phi.geom

Differentiable geometry package …

phi.jax

Jax integration …

phi.math

Alias for phiml.math …

phi.physics

Contains built-in physics functions, mainly for partial differential equations, such as incompressible fluids. The actual physics functions are …

phi.tf

TensorFlow integration …

phi.torch

PyTorch integration …

phi.vis

Visualization: plotting, interactive user interfaces …

Functions

def detect_backends() ‑> tuple

Registers all available backends and returns them. This includes only backends for which the minimal requirements are fulfilled.

Returns

tuple of phiml.backend.Backend

def set_logging_level(level='debug')

Sets the logging level for PhiFlow functions.

Args

level

Logging level, one of 'critical', 'fatal', 'error', 'warning', 'info', 'debug'

def verify()

Checks your configuration for potential problems and prints a summary.

To run verify without importing phi, run the script tests/verify.py included in the source distribution.