[MNT] add minimal dependency management utilities (#1628)

Adds minimal utilities for dependency management, to determine the packages installed.

This will be used later in isolation of dependencies that could be soft dependencies.

Instead of dumping the new utils in the current `utils` file, a new folder `utils` is added, in which the current `utils` is moved one level lower, and a `_dependencies` submodule is also added.

Author: fkiraly

.github/workflows/test.yml

@@ -103,7 +103,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v1
         with:
-          python-version: 3.8
+          python-version: 3.11
 
       - name: Cache pip
         uses: actions/cache@v2

.readthedocs.yml

@@ -9,7 +9,7 @@ version: 2
 # reference: https://docs.readthedocs.io/en/stable/config-file/v2.html#sphinx
 sphinx:
   configuration: docs/source/conf.py
-  fail_on_warning: true
+  # fail_on_warning: true
 
 # Build documentation with MkDocs
 #mkdocs:
@@ -21,6 +21,6 @@ formats:
 
 # Optionally set the version of Python and requirements required to build your docs
 python:
-  version: 3.8
+  version: 3.11
   install:
     - requirements: docs/requirements.txt

docs/requirements.txt

@@ -18,3 +18,4 @@ nbconvert >=6.3.0
 recommonmark >=0.7.1
 pytorch-optimizer >=2.5.1
 fastapi >0.80
+cpflows

docs/source/_templates/custom-module-template.rst

@@ -53,14 +53,14 @@
    {% endblock %}
 
 {% block modules %}
-{% if modules %}
+{% if all_modules %}
 .. rubric:: Modules
 
 .. autosummary::
    :toctree:
    :template: custom-module-template.rst
    :recursive:
-{% for item in modules %}
+{% for item in all_modules %}
    {{ item }}
 {%- endfor %}
 {% endif %}

docs/source/conf.py

@@ -176,3 +176,7 @@ def setup(app: Sphinx):
 intersphinx_mapping = {
     "sklearn": ("https://scikit-learn.org/stable/", None),
 }
+
+suppress_warnings = [
+    "autosummary.import_cycle",
+]

pyproject.toml

@@ -98,6 +98,7 @@ dev = [
   "pytest-dotenv>=0.5.2,<1.0.0",
   "tensorboard>=2.12.1,<3.0.0",
   "pandoc>=2.3,<3.0.0",
+  "cpflows",
 ]
 
 github-actions = ["pytest-github-actions-annotate-failures"]

pytorch_forecasting/metrics/_mqf2_utils.py

@@ -12,12 +12,13 @@ class DeepConvexNet(DeepConvexFlow):
     r"""
     Class that takes a partially input convex neural network (picnn)
     as input and equips it with functions of logdet
-    computation (both estimation and exact computation)
+    computation (both estimation and exact computation).
     This class is based on DeepConvexFlow of the CP-Flow
     repo (https://github.com/CW-Huang/CP-Flow)
     For details of the logdet estimator, see
     ``Convex potential flows: Universal probability distributions
     with optimal transport and convex optimization``
+
     Parameters
     ----------
     picnn
@@ -94,6 +95,7 @@ class SequentialNet(SequentialFlow):
     layers and provides energy score computation
     This class is based on SequentialFlow of the CP-Flow repo
     (https://github.com/CW-Huang/CP-Flow)
+
     Parameters
     ----------
     networks
@@ -116,6 +118,7 @@ def es_sample(self, hidden_state: torch.Tensor, dimension: int) -> torch.Tensor:
         """
         Auxiliary function for energy score computation
         Drawing samples conditioned on the hidden state
+
         Parameters
         ----------
         hidden_state
@@ -159,6 +162,7 @@ def energy_score(
         h_i is the hidden state associated with z_i,
         and es_num_samples is the number of samples drawn
         for each of w, w', w'' in energy score approximation
+
         Parameters
         ----------
         z
@@ -224,6 +228,7 @@ class MQF2Distribution(Distribution):
     Distribution class for the model MQF2 proposed in the paper
     ``Multivariate Quantile Function Forecaster``
     by Kan, Aubet, Januschowski, Park, Benidis, Ruthotto, Gasthaus
+
     Parameters
     ----------
     picnn
@@ -290,6 +295,7 @@ def stack_sliding_view(self, z: torch.Tensor) -> torch.Tensor:
         over the observations z
         Then, reshapes the observations into a 2-dimensional tensor for
         further computation
+
         Parameters
         ----------
         z
@@ -317,6 +323,7 @@ def log_prob(self, z: torch.Tensor) -> torch.Tensor:
         """
         Computes the log likelihood  log(g(z)) + logdet(dg(z)/dz),
         where g is the gradient of the picnn
+
         Parameters
         ----------
         z
@@ -346,6 +353,7 @@ def energy_score(self, z: torch.Tensor) -> torch.Tensor:
         h_i is the hidden state associated with z_i,
         and es_num_samples is the number of samples drawn
         for each of w, w', w'' in energy score approximation
+
         Parameters
         ----------
         z
@@ -370,14 +378,15 @@ def energy_score(self, z: torch.Tensor) -> torch.Tensor:
     def rsample(self, sample_shape: torch.Size = torch.Size()) -> torch.Tensor:
         """
         Generates the sample paths
+
         Parameters
         ----------
         sample_shape
             Shape of the samples
         Returns
         -------
         sample_paths
-            Tesnor of shape (batch_size, *sample_shape, prediction_length)
+            Tesnor of shape (batch_size, * sample_shape, prediction_length)
         """
 
         numel_batch = self.numel_batch
@@ -407,6 +416,7 @@ def rsample(self, sample_shape: torch.Size = torch.Size()) -> torch.Tensor:
     def quantile(self, alpha: torch.Tensor, hidden_state: Optional[torch.Tensor] = None) -> torch.Tensor:
         """
         Generates the predicted paths associated with the quantile levels alpha
+
         Parameters
         ----------
         alpha

pytorch_forecasting/utils/__init__.py

@@ -0,0 +1,51 @@
+"""
+PyTorch Forecasting package for timeseries forecasting with PyTorch.
+"""
+
+from pytorch_forecasting.utils._utils import (
+    InitialParameterRepresenterMixIn,
+    OutputMixIn,
+    TupleOutputMixIn,
+    apply_to_list,
+    autocorrelation,
+    concat_sequences,
+    create_mask,
+    detach,
+    get_embedding_size,
+    groupby_apply,
+    integer_histogram,
+    masked_op,
+    move_to_device,
+    padded_stack,
+    profile,
+    redirect_stdout,
+    repr_class,
+    to_list,
+    unpack_sequence,
+    unsqueeze_like,
+)
+
+__all__ = [
+    "InitialParameterRepresenterMixIn",
+    "OutputMixIn",
+    "TupleOutputMixIn",
+    "apply_to_list",
+    "autocorrelation",
+    "get_embedding_size",
+    "concat_sequences",
+    "create_mask",
+    "to_list",
+    "RecurrentNetwork",
+    "DecoderMLP",
+    "detach",
+    "masked_op",
+    "move_to_device",
+    "integer_histogram",
+    "groupby_apply",
+    "padded_stack",
+    "profile",
+    "redirect_stdout",
+    "repr_class",
+    "unpack_sequence",
+    "unsqueeze_like",
+]

pytorch_forecasting/utils/_dependencies.py

@@ -0,0 +1,40 @@
+"""Utilities for managing dependencies.
+
+Copied from sktime/skbase.
+"""
+
+from functools import lru_cache
+
+
+@lru_cache
+def _get_installed_packages_private():
+    """Get a dictionary of installed packages and their versions.
+
+    Same as _get_installed_packages, but internal to avoid mutating the lru_cache
+    by accident.
+    """
+    from importlib.metadata import distributions, version
+
+    dists = distributions()
+    package_names = {dist.metadata["Name"] for dist in dists}
+    package_versions = {pkg_name: version(pkg_name) for pkg_name in package_names}
+    # developer note:
+    # we cannot just use distributions naively,
+    # because the same top level package name may appear *twice*,
+    # e.g., in a situation where a virtual env overrides a base env,
+    # such as in deployment environments like databricks.
+    # the "version" contract ensures we always get the version that corresponds
+    # to the importable distribution, i.e., the top one in the sys.path.
+    return package_versions
+
+
+def _get_installed_packages():
+    """Get a dictionary of installed packages and their versions.
+
+    Returns
+    -------
+    dict : dictionary of installed packages and their versions
+        keys are PEP 440 compatible package names, values are package versions
+        MAJOR.MINOR.PATCH version format is used for versions, e.g., "1.2.3"
+    """
+    return _get_installed_packages_private().copy()