Merge branch 'master' into curve_average

.github/assistant.py

@@ -79,7 +79,7 @@ def set_min_torch_by_python(fpath: str = "requirements/base.txt") -> None:
             return
         with open(fpath) as fp:
             reqs = parse_requirements(fp.readlines())
-        pkg_ver = [p for p in reqs if p.name == "torch"][0]
+        pkg_ver = next(p for p in reqs if p.name == "torch")
         pt_ver = min([LooseVersion(v[1]) for v in pkg_ver.specs])
         pt_ver = max(LooseVersion(LUT_PYTHON_TORCH[py_ver]), pt_ver)
         with open(fpath) as fp:

.pre-commit-config.yaml

@@ -38,22 +38,22 @@ repos:
       - id: detect-private-key
 
   - repo: https://github.com/asottile/pyupgrade
-    rev: v3.9.0
+    rev: v3.14.0
     hooks:
       - id: pyupgrade
-        args: [--py38-plus]
+        args: ["--py38-plus"]
         name: Upgrade code
 
   - repo: https://github.com/codespell-project/codespell
-    rev: v2.2.5
+    rev: v2.2.6
     hooks:
       - id: codespell
         additional_dependencies: [tomli]
-        #args: ["--write-changes"]
+        args: ["--write-changes"]
         exclude: pyproject.toml
 
   - repo: https://github.com/crate-ci/typos
-    rev: v1.16.12
+    rev: v1.16.17
     hooks:
       - id: typos
         # empty to do not write fixes
@@ -68,13 +68,13 @@ repos:
         args: ["--in-place"]
 
   - repo: https://github.com/psf/black
-    rev: 23.7.0
+    rev: 23.9.1
     hooks:
       - id: black
         name: Format code
 
   - repo: https://github.com/executablebooks/mdformat
-    rev: 0.7.16
+    rev: 0.7.17
     hooks:
       - id: mdformat
         additional_dependencies:
@@ -130,7 +130,7 @@ repos:
       - id: text-unicode-replacement-char
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.0.277
+    rev: v0.0.292
     hooks:
       - id: ruff
         args: ["--fix"]

pyproject.toml

@@ -22,10 +22,8 @@ addopts = [
 #filterwarnings = ["error::FutureWarning"] # ToDo
 xfail_strict = true
 junit_duration_report = "call"
-
 [tool.coverage.report]
 exclude_lines = ["pragma: no cover", "pass"]
-
 [tool.coverage.run]
 parallel = true
 concurrency = "thread"
@@ -81,6 +79,7 @@ wil = "wil"
 
 
 [tool.ruff]
+target-version = "py38"
 line-length = 120
 # Enable Pyflakes `E` and `F` codes by default.
 select = [
@@ -122,6 +121,8 @@ ignore = [
     "S301",   # todo: `pickle` and modules that wrap it can be unsafe when used to deserialize untrusted data, possible security issue  # todo
     "S310",   # todo: Audit URL open for permitted schemes. Allowing use of `file:` or custom schemes is often unexpected.  # todo
     "B905",   # todo: `zip()` without an explicit `strict=` parameter
+    "PYI024",  # todo: Use `typing.NamedTuple` instead of `collections.namedtuple`
+    "PYI041",  # todo: Use `float` instead of `int | float``
 ]
 # Exclude a variety of commonly ignored directories.
 exclude = [

requirements/test.txt

@@ -1,7 +1,7 @@
 # NOTE: the upper bound for the package version is only set for CI stability, and it is dropped while installing this package
 #  in case you want to preserve/enforce restrictions on the latest compatible version, add "strict" as an in-line comment
 
-coverage ==7.3.1
+coverage ==7.3.2
 pytest ==7.4.2
 pytest-cov ==4.1.0
 pytest-doctestplus ==1.0.0
@@ -14,4 +14,4 @@ requests <=2.31.0
 fire <=0.5.0
 
 cloudpickle >1.3, <=2.2.1
-scikit-learn >=1.1.1, <1.3.1
+scikit-learn >=1.1.1, <1.4.0

src/torchmetrics/audio/__init__.py

@@ -41,16 +41,16 @@
 ]
 
 if _PESQ_AVAILABLE:
-    from torchmetrics.audio.pesq import PerceptualEvaluationSpeechQuality  # noqa: F401
+    from torchmetrics.audio.pesq import PerceptualEvaluationSpeechQuality
 
-    __all__.append("PerceptualEvaluationSpeechQuality")
+    __all__ += ["PerceptualEvaluationSpeechQuality"]
 
 if _PYSTOI_AVAILABLE:
-    from torchmetrics.audio.stoi import ShortTimeObjectiveIntelligibility  # noqa: F401
+    from torchmetrics.audio.stoi import ShortTimeObjectiveIntelligibility
 
-    __all__.append("ShortTimeObjectiveIntelligibility")
+    __all__ += ["ShortTimeObjectiveIntelligibility"]
 
 if _GAMMATONE_AVAILABLE and _TORCHAUDIO_AVAILABLE and _TORCHAUDIO_GREATER_EQUAL_0_10:
-    from torchmetrics.audio.srmr import SpeechReverberationModulationEnergyRatio  # noqa: F401
+    from torchmetrics.audio.srmr import SpeechReverberationModulationEnergyRatio
 
-    __all__.append("SpeechReverberationModulationEnergyRatio")
+    __all__ += ["SpeechReverberationModulationEnergyRatio"]

src/torchmetrics/audio/stoi.py

@@ -34,7 +34,7 @@ class ShortTimeObjectiveIntelligibility(Metric):
     The STOI-measure is intrusive, i.e., a function of the clean and degraded speech signals. STOI may be a good
     alternative to the speech intelligibility index (SII) or the speech transmission index (STI), when you are
     interested in the effect of nonlinear processing to noisy speech, e.g., noise reduction, binary masking algorithms,
-    on speech intelligibility. Description taken from  `Cees Taal's website`_ and for further defails see `STOI ref1`_
+    on speech intelligibility. Description taken from  `Cees Taal's website`_ and for further details see `STOI ref1`_
     and `STOI ref2`_.
 
     This metric is a wrapper for the `pystoi package`_. As the implementation backend implementation only supports

src/torchmetrics/classification/accuracy.py

@@ -49,7 +49,8 @@ class BinaryAccuracy(BinaryStatScores):
           If ``multidim_average`` is set to ``samplewise``, the metric returns ``(N,)`` vector consisting of a scalar
           value per sample.
 
-    Additional dimension ``...`` will be flattened into the batch dimension.
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
 
     Args:
         threshold: Threshold for transforming probability to binary {0,1} predictions
@@ -176,6 +177,9 @@ class MulticlassAccuracy(MulticlassStatScores):
               - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
               - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_classes: Integer specifying the number of classes
         average:
@@ -325,6 +329,9 @@ class MultilabelAccuracy(MultilabelStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions

src/torchmetrics/classification/exact_match.py

@@ -54,14 +54,16 @@ class MulticlassExactMatch(Metric):
       probabilities/logits into an int tensor.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mcem`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``multidim_average`` argument:
 
         - If ``multidim_average`` is set to ``global`` the output will be a scalar tensor
         - If ``multidim_average`` is set to ``samplewise`` the output will be a tensor of shape ``(N,)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_classes: Integer specifying the number of labels
         multidim_average:
@@ -206,14 +208,16 @@ class MultilabelExactMatch(Metric):
       sigmoid per element. Additionally, we convert to int tensor with thresholding using the value in ``threshold``.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, C, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mlem`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``multidim_average`` argument:
 
         - If ``multidim_average`` is set to ``global`` the output will be a scalar tensor
         - If ``multidim_average`` is set to ``samplewise`` the output will be a tensor of shape ``(N,)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions

src/torchmetrics/classification/f_beta.py

@@ -66,6 +66,9 @@ class BinaryFBetaScore(BinaryStatScores):
         - If ``multidim_average`` is set to ``samplewise`` the output will be a tensor of shape ``(N,)`` consisting of
           a scalar value per sample.
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         beta: Weighting between precision and recall in calculation. Setting to 1 corresponds to equal weight
         threshold: Threshold for transforming probability to binary {0,1} predictions
@@ -202,7 +205,6 @@ class MulticlassFBetaScore(MulticlassStatScores):
       probabilities/logits into an int tensor.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mcfbs`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``average`` and
@@ -218,6 +220,9 @@ class MulticlassFBetaScore(MulticlassStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         beta: Weighting between precision and recall in calculation. Setting to 1 corresponds to equal weight
         num_classes: Integer specifying the number of classes
@@ -382,7 +387,6 @@ class MultilabelFBetaScore(MultilabelStatScores):
       per element. Additionally, we convert to int tensor with thresholding using the value in ``threshold``.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, C, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mlfbs`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``average`` and
@@ -398,6 +402,9 @@ class MultilabelFBetaScore(MultilabelStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         beta: Weighting between precision and recall in calculation. Setting to 1 corresponds to equal weight
         num_labels: Integer specifying the number of labels
@@ -566,6 +573,9 @@ class BinaryF1Score(BinaryFBetaScore):
         - If ``multidim_average`` is set to ``samplewise``, the metric returns ``(N,)`` vector consisting of a scalar
           value per sample.
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         threshold: Threshold for transforming probability to binary {0,1} predictions
         multidim_average:
@@ -706,6 +716,9 @@ class MulticlassF1Score(MulticlassFBetaScore):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         preds: Tensor with predictions
         target: Tensor with true labels
@@ -876,6 +889,9 @@ class MultilabelF1Score(MultilabelFBetaScore):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)```
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions

src/torchmetrics/classification/hamming.py

@@ -58,6 +58,9 @@ class BinaryHammingDistance(BinaryStatScores):
         - If ``multidim_average`` is set to ``samplewise``, the metric returns ``(N,)`` vector consisting of a
           scalar value per sample.
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         threshold: Threshold for transforming probability to binary {0,1} predictions
         multidim_average:
@@ -171,7 +174,6 @@ class MulticlassHammingDistance(MulticlassStatScores):
       probabilities/logits into an int tensor.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mchd`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``average`` and
@@ -187,6 +189,9 @@ class MulticlassHammingDistance(MulticlassStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_classes: Integer specifying the number of classes
         average:
@@ -324,7 +329,6 @@ class MultilabelHammingDistance(MultilabelStatScores):
       ``threshold``.
     - ``target`` (:class:`~torch.Tensor`): An int tensor of shape ``(N, C, ...)``.
 
-
     As output to ``forward`` and ``compute`` the metric returns the following output:
 
     - ``mlhd`` (:class:`~torch.Tensor`): A tensor whose returned shape depends on the ``average`` and
@@ -340,6 +344,9 @@ class MultilabelHammingDistance(MultilabelStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions

src/torchmetrics/classification/precision_recall.py

@@ -57,6 +57,9 @@ class BinaryPrecision(BinaryStatScores):
       value. If ``multidim_average`` is set to ``samplewise``, the metric returns ``(N,)`` vector consisting of a
       scalar value per sample.
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         threshold: Threshold for transforming probability to binary {0,1} predictions
         multidim_average:
@@ -187,6 +190,9 @@ class MulticlassPrecision(MulticlassStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_classes: Integer specifying the number of classes
         average:
@@ -340,6 +346,9 @@ class MultilabelPrecision(MultilabelStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions
@@ -479,6 +488,9 @@ class BinaryRecall(BinaryStatScores):
       value. If ``multidim_average`` is set to ``samplewise``, the metric returns ``(N,)`` vector consisting of
       a scalar value per sample.
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         threshold: Threshold for transforming probability to binary {0,1} predictions
         multidim_average:
@@ -608,6 +620,9 @@ class MulticlassRecall(MulticlassStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_classes: Integer specifying the number of classes
         average:
@@ -760,6 +775,9 @@ class MultilabelRecall(MultilabelStatScores):
           - If ``average='micro'/'macro'/'weighted'``, the shape will be ``(N,)``
           - If ``average=None/'none'``, the shape will be ``(N, C)``
 
+    If ``multidim_average`` is set to ``samplewise`` we expect at least one additional dimension ``...`` to be present,
+    which the reduction will then be applied over instead of the sample dimension ``N``.
+
     Args:
         num_labels: Integer specifying the number of labels
         threshold: Threshold for transforming probability to binary (0,1) predictions