Bumping numpy and jax versions.

.pre-commit-config.yaml

@@ -12,7 +12,7 @@ default_language_version:
   python: python3
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v3.2.0
+    rev: v5.0.0
     hooks:
       - id: check-added-large-files
         # Default is 500 which is too strict on the notebooks.
@@ -31,7 +31,7 @@ repos:
         # exclude autogenerated files
         exclude: /README\.rst$|\.pot?$
   - repo: https://github.com/psf/black
-    rev: 22.3.0 # Make sure to use the same tag/version as specified in formatting-requirements.txt
+    rev: 25.1.0 # Make sure to use the same tag/version as specified in formatting-requirements.txt
     hooks:
       - id: black
         language_version: python3
@@ -40,7 +40,7 @@ repos:
     hooks:
       - id: isort
   - repo: https://github.com/myint/docformatter
-    rev: v1.3.1
+    rev: v1.7.7
     hooks:
       - id: docformatter
         args: ['--in-place',

benchmarks/.gitignore

@@ -1,2 +1,2 @@
 .asv/
-probnum/
+probnum/

benchmarks/filtsmooth.py

@@ -1,4 +1,5 @@
 """Benchmarks for Gaussian filtering."""
+
 import functools
 
 import numpy as np

benchmarks/ivpsolvers.py

@@ -1,4 +1,5 @@
 """Benchmarks for probabilistic IVP solvers."""
+
 import numpy as np
 
 from probnum.diffeq import probsolve_ivp

benchmarks/linearsolvers.py

@@ -1,4 +1,5 @@
 """Benchmarks for linear solvers."""
+
 import numpy as np
 
 from probnum import linops, problems, randvars

benchmarks/randprocs.py

@@ -1,4 +1,5 @@
 """Benchmarks for random processes."""
+
 import numpy as np
 import scipy.stats
 

docs/notebook-requirements.txt

@@ -1,5 +1,5 @@
 # Jupyter notebooks
-ipython>=8.9.0<8.11.0
+ipython>=8.9.0,<8.11.0
 jupyter
 
 # Notebook conversion

docs/source/development/adding_to_the_api_documentation.rst

@@ -17,7 +17,7 @@ detail its functionality. This package uses the `NumPy docstring
 format <https://numpydoc.readthedocs.io/en/latest/format.html#numpydoc-docstring-guide%3E>`__.
 As a rule, all functions which are exposed to the user *must* have
 appropriate docstrings. Below is an example of a docstring for a
-probabilistic numerical method 
+probabilistic numerical method
 named ``problinsolve`` defined in ``_problinsolve.py`` in the ``linalg`` subpackage.
 
 .. literalinclude:: ../../../src/probnum/linalg/_problinsolve.py
@@ -103,7 +103,7 @@ documentation. You can configure Sphinx itself or its extensions in the
 .. code:: ipython3
 
     from IPython.display import Image
-    
+
     display(Image(filename="../assets/img/developer_guides/sphinx_logo.png", embed=True))
 
 

docs/source/development/creating_an_example_notebook.rst

@@ -10,8 +10,8 @@ follow the instructions below.
 .. code:: ipython3
 
     from IPython.display import Image
-    
-    display(Image(url='https://upload.wikimedia.org/wikipedia/commons/thumb/3/38/Jupyter_logo.svg/1200px-Jupyter_logo.svg.png', 
+
+    display(Image(url='https://upload.wikimedia.org/wikipedia/commons/thumb/3/38/Jupyter_logo.svg/1200px-Jupyter_logo.svg.png',
                   width=250, embed=True))
 
 
@@ -44,12 +44,12 @@ of the notebook should contain the following code snippet.
     # Make inline plots vector graphics instead of raster graphics
     %matplotlib inline
     from IPython.display import set_matplotlib_formats
-    
+
     set_matplotlib_formats("pdf", "svg")
-    
+
     # Plotting
     import matplotlib.pyplot as plt
-    
+
     plt.style.use("../probnum.mplstyle")
 
 This snippet defines the style of plots generated by Matplotlib. For

docs/source/development/implementing_a_probnum_method.ipynb

@@ -728,7 +728,7 @@
     "# %load -s function_evaluation quadopt_example/observation_operators\n",
     "def function_evaluation(\n",
     "    fun: Callable[[FloatLike], FloatLike], action: FloatLike\n",
-    ") -> np.float_:\n",
+    ") -> np.float64:\n",
     "    \"\"\"Observe a (noisy) function evaluation of the quadratic objective.\n",
     "\n",
     "    Parameters\n",

docs/source/development/quadopt_example/_probsolve_qp.py

@@ -133,7 +133,7 @@ def probsolve_qp(
 
 
 def _choose_prior(
-    fun_params0: Union[randvars.RandomVariable, np.ndarray, None]
+    fun_params0: Union[randvars.RandomVariable, np.ndarray, None],
 ) -> randvars.RandomVariable:
     """Initialize the prior distribution over the parameters.
 

docs/source/development/quadopt_example/observation_operators.py

@@ -10,7 +10,7 @@
 
 def function_evaluation(
     fun: Callable[[FloatLike], FloatLike], action: FloatLike
-) -> np.float_:
+) -> np.float64:
     """Observe a (noisy) function evaluation of the quadratic objective.
 
     Parameters

docs/source/development/unit_testing.rst

@@ -60,7 +60,7 @@ check out the following guides:
 .. code:: ipython3
 
     from IPython.display import Image
-    
+
     display(Image(url='https://code.visualstudio.com/assets/docs/python/testing/test-results.png', embed=True))
 
 
@@ -86,7 +86,7 @@ test. Here is a simple example.
 
     def add(x, y):
         return x + y
-    
+
     def test_symmetric():
         assert add(3, 5) == add(5, 3)
 
@@ -96,7 +96,7 @@ You can do so by parametrizing tests via a decorator.
 .. code:: ipython3
 
     import pytest
-    
+
     @pytest.mark.parametrize("x,y", [(3, 5), (-1, 1), (0, 1.2)])
     def test_symmetric(x, y):
         assert add(x, y) == add(y, x)

docs/source/tutorials/linops/linear_operators_quickstart.ipynb

@@ -380,7 +380,7 @@
     "    return np.roll(v, 1)\n",
     "\n",
     "n = 5\n",
-    "P_op = LambdaLinearOperator(shape=(n, n), dtype=np.float_, matmul=mv)\n",
+    "P_op = LambdaLinearOperator(shape=(n, n), dtype=np.float64, matmul=mv)\n",
     "x = np.arange(0., n, 1)\n",
     "\n",
     "P_op"
@@ -512,7 +512,7 @@
     "def mv(v):\n",
     "    return v[:n-1]\n",
     "\n",
-    "Pr = LambdaLinearOperator(shape=(n-1, n), dtype=np.float_, matmul=mv)\n",
+    "Pr = LambdaLinearOperator(shape=(n-1, n), dtype=np.float64, matmul=mv)\n",
     "\n",
     "# Apply the operator to the 3D normal random variable\n",
     "rv_projected = Pr @ rv"

setup.py

@@ -9,7 +9,7 @@
 # `pyproject.toml`, once optional dependencies defined there can reference one another
 extras_require = dict()
 extras_require["jax"] = [
-    "jax[cpu]<0.4.20; platform_system!='Windows'",
+    "jax[cpu]>=0.4.26; platform_system!='Windows'",
 ]
 extras_require["zoo"] = [
     "tqdm>=4.0",

src/probnum/.gitignore

@@ -1 +1 @@
-_version.py
+_version.py

src/probnum/_config.py

@@ -1,4 +1,4 @@
-"""ProbNum library configuration"""
+"""ProbNum library configuration."""
 
 import contextlib
 import dataclasses
@@ -165,6 +165,7 @@ def register(self, key: str, default_value: Any, description: str) -> None:
     ),
 ]
 
+
 # ... and register the default configuration options.
 def _register_defaults():
     for key, default_value, descr in _DEFAULT_CONFIG_OPTIONS:

src/probnum/conftest.py

@@ -9,7 +9,7 @@
 @pytest.fixture(autouse=True)
 def autoimport_packages(doctest_namespace):  # pylint: disable=missing-any-param-doc
     """This fixture 'imports' standard packages automatically in order to avoid
-    boilerplate code in doctests"""
+    boilerplate code in doctests."""
 
     doctest_namespace["pn"] = pn
     doctest_namespace["np"] = np

src/probnum/diffeq/_odesolution.py

@@ -2,8 +2,8 @@
 
 This object is returned by ODESolver.solve().
 
-Provides dense output (by being callable), is sliceable,
-and collects the time-grid as well as the discrete-time solution.
+Provides dense output (by being callable), is sliceable, and collects the time-grid as
+well as the discrete-time solution.
 """
 
 from typing import Optional

src/probnum/diffeq/_probsolve_ivp.py

@@ -267,7 +267,6 @@ def probsolve_ivp(
      [0.97]
      [0.98]
      [0.99]]
-
     """
 
     # Create IVP object

src/probnum/diffeq/odefilter/_odefilter.py

@@ -56,7 +56,6 @@ class ODEFilter(_odesolver.ODESolver):
     .. [2] Bosch, N., and Hennig, P., and Tronarp, F..
         Calibrated adaptive probabilistic ODE solvers.
         2021.
-
     """
 
     def __init__(
@@ -196,7 +195,6 @@ def attempt_step(self, state, dt):
         5. With the results of either 4.1. or 4.2. (which both return a predicted RV and
         a measured RV), we finally compute the Kalman update and return the result.
         Recall that the error estimate has been computed in the third step.
-
         """
 
         # Read off system matrices; required for calibration / error estimation

src/probnum/diffeq/odefilter/approx_strategies/_ek.py

@@ -2,6 +2,7 @@
 
 Make an intractable information operator tractable with local linearization.
 """
+
 from typing import Optional
 
 import numpy as np

src/probnum/diffeq/odefilter/init_routines/_autodiff.py

@@ -11,14 +11,14 @@
 # pylint: disable="import-outside-toplevel"
 try:
     import jax
-    from jax.config import config
+    from jax import config
     from jax.experimental.jet import jet
     import jax.numpy as jnp
 
     config.update("jax_enable_x64", True)
 
     JAX_IS_AVAILABLE = True
-except ImportError as JAX_IMPORT_ERROR:
+except ImportError:
     JAX_IS_AVAILABLE = False
     JAX_IMPORT_ERROR_MSG = (
         "Cannot perform Jax-based initialization without the optional "
@@ -31,7 +31,7 @@ class _AutoDiffBase(InitializationRoutine):
     def __init__(self):
 
         if not JAX_IS_AVAILABLE:
-            raise ImportError(JAX_IMPORT_ERROR_MSG) from JAX_IMPORT_ERROR
+            raise ImportError(JAX_IMPORT_ERROR_MSG)
 
         super().__init__(is_exact=True, requires_jax=True)
 
@@ -70,12 +70,12 @@ def _make_autonomous(self, *, ivp):
         Turn the ODE into a format that is more convenient to handle with automatic
         differentiation. This has no effect on the ODE itself. It is purely internal.
         """
-        y0_autonomous = jnp.concatenate([ivp.y0, jnp.array([ivp.t0])])
+        y0_autonomous = jnp.concatenate([ivp.y0, jnp.asarray([ivp.t0])])
 
         def f_autonomous(y):
             x, t = y[:-1], y[-1]
             fx = ivp.f(t, x)
-            return jnp.concatenate([fx, jnp.array([1.0])])
+            return jnp.concatenate([fx, jnp.asarray([1.0])])
 
         return f_autonomous, y0_autonomous
 

src/probnum/diffeq/odefilter/init_routines/_interface.py

@@ -1,6 +1,5 @@
 """Interface for initialization routines."""
 
-
 import abc
 
 from probnum import problems, randprocs, randvars

src/probnum/diffeq/odefilter/init_routines/_non_probabilistic_fit.py

@@ -1,7 +1,6 @@
 """Initialization routines based on fitting the prior to (a few steps of a) non-
 probabilistc solver."""
 
-
 from typing import Optional
 
 import numpy as np

src/probnum/diffeq/odefilter/init_routines/_stack.py

@@ -1,6 +1,5 @@
 """Stacking-based initialization routines."""
 
-
 import numpy as np
 
 from probnum import problems, randprocs, randvars

src/probnum/diffeq/odefilter/utils/_problem_utils.py

@@ -10,6 +10,7 @@
 
 __all__ = ["ivp_to_regression_problem"]
 
+
 # The ODE information operator is not optional, because in order to create it
 # one needs to know the order of the algorithm that is desired
 # (i.e. num_prior_derivatives). Since this is a weird input for the function, it seems

src/probnum/diffeq/perturbed/scipy_wrapper/_wrapped_scipy_odesolution.py

@@ -1,4 +1,5 @@
 """Make a ProbNum ODE solution out of a scipy ODE solution."""
+
 import numpy as np
 from scipy.integrate._ivp.common import OdeSolution
 

src/probnum/diffeq/perturbed/scipy_wrapper/_wrapped_scipy_solver.py

@@ -3,6 +3,7 @@
 Dense-output can not be used for DOP853, if you use other RK-methods, make sure, that
 the current implementation works for them.
 """
+
 import numpy as np
 from scipy.integrate._ivp import rk
 from scipy.integrate._ivp.common import OdeSolution

src/probnum/diffeq/perturbed/step/_perturbation_functions.py

@@ -1,4 +1,5 @@
 """Perturbation functions to perturb the stepsize."""
+
 from typing import Optional, Union
 
 import numpy as np

src/probnum/diffeq/stepsize/_propose_firststep.py

@@ -6,8 +6,8 @@
 def propose_firststep(ivp):
     """Propose a suitable first step-size that can be taken by an ODE solver.
 
-    This function implements a lazy version of the algorithm on p. 169
-    of Hairer, Wanner, Norsett.
+    This function implements a lazy version of the algorithm on p. 169 of Hairer,
+    Wanner, Norsett.
     """
     norm_y0 = np.linalg.norm(ivp.y0)
     norm_dy0 = np.linalg.norm(ivp.f(ivp.t0, ivp.y0))

src/probnum/filtsmooth/__init__.py

@@ -1,11 +1,11 @@
 """Bayesian Filtering and Smoothing.
 
-This package provides different kinds of Bayesian filters and smoothers
-which estimate the distribution over observed and hidden variables in a
-sequential model. The two operations differ by what information they
-use. Filtering considers all observations up to a given point, while
-smoothing takes the entire set of observations into account.
+This package provides different kinds of Bayesian filters and smoothers which estimate
+the distribution over observed and hidden variables in a sequential model. The two
+operations differ by what information they use. Filtering considers all observations up
+to a given point, while smoothing takes the entire set of observations into account.
 """
+
 from . import gaussian, particle, utils
 from ._bayesfiltsmooth import BayesFiltSmooth
 from ._kalman_filter_smoother import filter_kalman, smooth_rts

src/probnum/filtsmooth/_timeseriesposterior.py

@@ -13,9 +13,8 @@
 DenseOutputValueType = Union[randvars.RandomVariable, randvars._RandomVariableList]
 """Output type of interpolation.
 
-Dense evaluation of a TimeSeriesPosterior returns a RandomVariable
-if evaluated at a single location,
-and a _RandomVariableList if evaluated at an array of locations.
+Dense evaluation of a TimeSeriesPosterior returns a RandomVariable if evaluated at a
+single location, and a _RandomVariableList if evaluated at an array of locations.
 """
 
 

src/probnum/filtsmooth/gaussian/_kalman.py

@@ -1,6 +1,5 @@
 """Gaussian filtering and smoothing."""
 
-
 from typing import Iterable, Optional, Union
 
 import numpy as np