Merge branch 'master' into fix/segmentation-meaniou

CHANGELOG.md

@@ -48,6 +48,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Fixed wrong aggregation in `segmentation.MeanIoU` ([#2698](https://github.com/Lightning-AI/torchmetrics/pull/2698))
 
 
+- Fixed compatibility of audio domain with new `scipy` ([#2733](https://github.com/Lightning-AI/torchmetrics/pull/2733))
+
+
 - Fixed how `prefix`/`postfix` works in `MultitaskWrapper` ([#2722](https://github.com/Lightning-AI/torchmetrics/pull/2722))
 
 

src/torchmetrics/audio/__init__.py

@@ -28,10 +28,18 @@
     _ONNXRUNTIME_AVAILABLE,
     _PESQ_AVAILABLE,
     _PYSTOI_AVAILABLE,
+    _SCIPI_AVAILABLE,
     _TORCHAUDIO_AVAILABLE,
     _TORCHAUDIO_GREATER_EQUAL_0_10,
 )
 
+if _SCIPI_AVAILABLE:
+    import scipy.signal
+
+    # back compatibility patch due to SMRMpy using scipy.signal.hamming
+    if not hasattr(scipy.signal, "hamming"):
+        scipy.signal.hamming = scipy.signal.windows.hamming
+
 __all__ = [
     "PermutationInvariantTraining",
     "ScaleInvariantSignalDistortionRatio",

src/torchmetrics/functional/audio/__init__.py

@@ -28,10 +28,18 @@
     _ONNXRUNTIME_AVAILABLE,
     _PESQ_AVAILABLE,
     _PYSTOI_AVAILABLE,
+    _SCIPI_AVAILABLE,
     _TORCHAUDIO_AVAILABLE,
     _TORCHAUDIO_GREATER_EQUAL_0_10,
 )
 
+if _SCIPI_AVAILABLE:
+    import scipy.signal
+
+    # back compatibility patch due to SMRMpy using scipy.signal.hamming
+    if not hasattr(scipy.signal, "hamming"):
+        scipy.signal.hamming = scipy.signal.windows.hamming
+
 __all__ = [
     "permutation_invariant_training",
     "pit_permutate",

src/torchmetrics/functional/nominal/__init__.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 from torchmetrics.functional.nominal.cramers import cramers_v, cramers_v_matrix
 from torchmetrics.functional.nominal.fleiss_kappa import fleiss_kappa
 from torchmetrics.functional.nominal.pearson import (
@@ -19,6 +20,14 @@
 )
 from torchmetrics.functional.nominal.theils_u import theils_u, theils_u_matrix
 from torchmetrics.functional.nominal.tschuprows import tschuprows_t, tschuprows_t_matrix
+from torchmetrics.utilities.imports import _SCIPI_AVAILABLE
+
+if _SCIPI_AVAILABLE:
+    import scipy.signal
+
+    # back compatibility patch due to SMRMpy using scipy.signal.hamming
+    if not hasattr(scipy.signal, "hamming"):
+        scipy.signal.hamming = scipy.signal.windows.hamming
 
 __all__ = [
     "cramers_v",

src/torchmetrics/nominal/__init__.py

@@ -11,11 +11,20 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 from torchmetrics.nominal.cramers import CramersV
 from torchmetrics.nominal.fleiss_kappa import FleissKappa
 from torchmetrics.nominal.pearson import PearsonsContingencyCoefficient
 from torchmetrics.nominal.theils_u import TheilsU
 from torchmetrics.nominal.tschuprows import TschuprowsT
+from torchmetrics.utilities.imports import _SCIPI_AVAILABLE
+
+if _SCIPI_AVAILABLE:
+    import scipy.signal
+
+    # back compatibility patch due to SMRMpy using scipy.signal.hamming
+    if not hasattr(scipy.signal, "hamming"):
+        scipy.signal.hamming = scipy.signal.windows.hamming
 
 __all__ = [
     "CramersV",

src/torchmetrics/utilities/imports.py

@@ -64,6 +64,7 @@
 _MECAB_KO_DIC_AVAILABLE = RequirementCache("mecab_ko_dic")
 _IPADIC_AVAILABLE = RequirementCache("ipadic")
 _SENTENCEPIECE_AVAILABLE = RequirementCache("sentencepiece")
+_SCIPI_AVAILABLE = RequirementCache("scipy")
 _SKLEARN_GREATER_EQUAL_1_3 = RequirementCache("scikit-learn>=1.3.0")
 
 _LATEX_AVAILABLE: bool = shutil.which("latex") is not None

tests/unittests/classification/test_stat_scores.py

@@ -582,8 +582,8 @@ def test_support_for_int():
     """See issue: https://github.com/Lightning-AI/torchmetrics/issues/1970."""
     seed_all(42)
     metric = MulticlassStatScores(num_classes=4, average="none", multidim_average="samplewise", ignore_index=0)
-    prediction = torch.randint(low=0, high=4, size=(1, 224, 224)).to(torch.uint8)
-    label = torch.randint(low=0, high=4, size=(1, 224, 224)).to(torch.uint8)
+    prediction = torch.randint(low=0, high=4, size=(1, 50, 50)).to(torch.uint8)
+    label = torch.randint(low=0, high=4, size=(1, 50, 50)).to(torch.uint8)
     score = metric(preds=prediction, target=label)
     assert score.shape == (1, 4, 5)