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

Expand source code

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`
    """
    from phiml.backend._backend import init_backend
    try:
        init_backend('jax')
    except ImportError:
        pass
    try:
        init_backend('torch')
    except ImportError:
        pass
    try:
        init_backend('tensorflow')
    except ImportError:
        pass
    from phiml.backend import BACKENDS
    return tuple([b for b in BACKENDS if b.name not in {'Python', 'object'}])

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')

Expand source code

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'`
    """
    from phiml.backend import ML_LOGGER
    ML_LOGGER.setLevel(level.upper())

Sets the logging level for PhiFlow functions.

Args

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

def verify()

Expand source code

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.
    """
    import sys
    from ._troubleshoot import assert_minimal_config, troubleshoot
    try:
        assert_minimal_config()
    except AssertionError as fail_err:
        print("\n".join(fail_err.args), file=sys.stderr)
        return
    print(troubleshoot())

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.