Prune deprecated: audio

CHANGELOG.md

@@ -25,6 +25,19 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Removed deprecated functions, and warnings in Text ([#773](https://github.com/PyTorchLightning/metrics/pull/773))
   * `functional.wer`
   * `WER`
+- Removed deprecated functions, and warnings in Audio ([#805](https://github.com/PyTorchLightning/metrics/pull/805))
+  * `PESQ`
+  * `PIT`
+  * `SDR`
+  * `SNR`
+  * `STOI`
+  * `functional.audio.pesq`
+  * `functional.audio.pit`
+  * `functional.audio.sdr`
+  * `functional.audio.snr`
+  * `functional.audio.stoi`
+  * `functional.audio.si_sdr`
+  * `functional.audio.si_snr`
 
 
 ### Fixed

docs/source/references/functional.rst

@@ -41,7 +41,7 @@ scale_invariant_signal_distortion_ratio [func]
 scale_invariant_signal_noise_ratio [func]
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-.. autofunction:: torchmetrics.functional.si_snr
+.. autofunction:: torchmetrics.functional.scale_invariant_signal_noise_ratio
     :noindex:
 
 

docs/source/references/modules.rst

@@ -69,12 +69,6 @@ the metric will be computed over the ``time`` dimension.
     >>> snr = SignalNoiseRatio()
     >>> snr(preds, target)
     tensor(16.1805)
-    >>> from torchmetrics import SNR
-    >>> target = torch.tensor([3.0, -0.5, 2.0, 7.0])
-    >>> preds = torch.tensor([2.5, 0.0, 2.0, 8.0])
-    >>> snr = SignalNoiseRatio()
-    >>> snr(preds, target)
-    tensor(16.1805)
 
 PerceptualEvaluationSpeechQuality
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

tests/audio/test_pit.py

@@ -24,7 +24,11 @@
 from tests.helpers import seed_all
 from tests.helpers.testers import BATCH_SIZE, NUM_BATCHES, MetricTester
 from torchmetrics.audio import PermutationInvariantTraining
-from torchmetrics.functional import permutation_invariant_training, scale_invariant_signal_distortion_ratio, snr
+from torchmetrics.functional import (
+    permutation_invariant_training,
+    scale_invariant_signal_distortion_ratio,
+    signal_noise_ratio,
+)
 from torchmetrics.utilities.imports import _TORCH_GREATER_EQUAL_1_6
 
 seed_all(42)
@@ -96,7 +100,7 @@ def _average_metric(preds: Tensor, target: Tensor, metric_func: Callable) -> Ten
     return metric_func(preds, target)[0].mean()
 
 
-snr_pit_scipy = partial(naive_implementation_pit_scipy, metric_func=snr, eval_func="max")
+snr_pit_scipy = partial(naive_implementation_pit_scipy, metric_func=signal_noise_ratio, eval_func="max")
 si_sdr_pit_scipy = partial(
     naive_implementation_pit_scipy, metric_func=scale_invariant_signal_distortion_ratio, eval_func="max"
 )
@@ -105,9 +109,9 @@ def _average_metric(preds: Tensor, target: Tensor, metric_func: Callable) -> Ten
 @pytest.mark.parametrize(
     "preds, target, sk_metric, metric_func, eval_func",
     [
-        (inputs1.preds, inputs1.target, snr_pit_scipy, snr, "max"),
+        (inputs1.preds, inputs1.target, snr_pit_scipy, signal_noise_ratio, "max"),
         (inputs1.preds, inputs1.target, si_sdr_pit_scipy, scale_invariant_signal_distortion_ratio, "max"),
-        (inputs2.preds, inputs2.target, snr_pit_scipy, snr, "max"),
+        (inputs2.preds, inputs2.target, snr_pit_scipy, signal_noise_ratio, "max"),
         (inputs2.preds, inputs2.target, si_sdr_pit_scipy, scale_invariant_signal_distortion_ratio, "max"),
     ],
 )
@@ -166,19 +170,19 @@ def test_pit_half_gpu(self, preds, target, sk_metric, metric_func, eval_func):
 
 
 def test_error_on_different_shape() -> None:
-    metric = PermutationInvariantTraining(snr, "max")
+    metric = PermutationInvariantTraining(signal_noise_ratio, "max")
     with pytest.raises(RuntimeError, match="Predictions and targets are expected to have the same shape"):
         metric(torch.randn(3, 3, 10), torch.randn(3, 2, 10))
 
 
 def test_error_on_wrong_eval_func() -> None:
-    metric = PermutationInvariantTraining(snr, "xxx")
+    metric = PermutationInvariantTraining(signal_noise_ratio, "xxx")
     with pytest.raises(ValueError, match='eval_func can only be "max" or "min"'):
         metric(torch.randn(3, 3, 10), torch.randn(3, 3, 10))
 
 
 def test_error_on_wrong_shape() -> None:
-    metric = PermutationInvariantTraining(snr, "max")
+    metric = PermutationInvariantTraining(signal_noise_ratio, "max")
     with pytest.raises(ValueError, match="Inputs must be of shape *"):
         metric(torch.randn(3), torch.randn(3))
 

torchmetrics/__init__.py

@@ -14,11 +14,6 @@
 from torchmetrics import functional  # noqa: E402
 from torchmetrics.aggregation import CatMetric, MaxMetric, MeanMetric, MinMetric, SumMetric  # noqa: E402
 from torchmetrics.audio import (  # noqa: E402
-    PIT,
-    SDR,
-    SI_SDR,
-    SI_SNR,
-    SNR,
     PermutationInvariantTraining,
     ScaleInvariantSignalDistortionRatio,
     ScaleInvariantSignalNoiseRatio,
@@ -151,7 +146,6 @@
     "PearsonCorrcoef",
     "PearsonCorrCoef",
     "PermutationInvariantTraining",
-    "PIT",
     "Precision",
     "PrecisionRecallCurve",
     "PSNR",
@@ -168,14 +162,10 @@
     "RetrievalRPrecision",
     "ROC",
     "SacreBLEUScore",
-    "SDR",
     "SignalDistortionRatio",
     "ScaleInvariantSignalDistortionRatio",
-    "SI_SDR",
-    "SI_SNR",
     "ScaleInvariantSignalNoiseRatio",
     "SignalNoiseRatio",
-    "SNR",
     "SpearmanCorrcoef",
     "SpearmanCorrCoef",
     "Specificity",

torchmetrics/audio/__init__.py

@@ -11,11 +11,9 @@
 # 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.audio.pit import PIT, PermutationInvariantTraining  # noqa: F401
-from torchmetrics.audio.sdr import SDR, ScaleInvariantSignalDistortionRatio, SignalDistortionRatio  # noqa: F401
-from torchmetrics.audio.si_sdr import SI_SDR  # noqa: F401
-from torchmetrics.audio.si_snr import SI_SNR  # noqa: F401
-from torchmetrics.audio.snr import SNR, ScaleInvariantSignalNoiseRatio, SignalNoiseRatio  # noqa: F401
+from torchmetrics.audio.pit import PermutationInvariantTraining  # noqa: F401
+from torchmetrics.audio.sdr import ScaleInvariantSignalDistortionRatio, SignalDistortionRatio  # noqa: F401
+from torchmetrics.audio.snr import ScaleInvariantSignalNoiseRatio, SignalNoiseRatio  # noqa: F401
 from torchmetrics.utilities.imports import _PESQ_AVAILABLE, _PYSTOI_AVAILABLE
 
 if _PESQ_AVAILABLE:

torchmetrics/audio/pesq.py

@@ -13,15 +13,13 @@
 # limitations under the License.
 from typing import Any, Callable, Optional
 
-from deprecate import deprecated, void
 from torch import Tensor, tensor
 
 from torchmetrics.functional.audio.pesq import perceptual_evaluation_speech_quality
 from torchmetrics.metric import Metric
-from torchmetrics.utilities import _future_warning
 from torchmetrics.utilities.imports import _PESQ_AVAILABLE
 
-__doctest_requires__ = {("PerceptualEvaluationSpeechQuality", "PESQ"): ["pesq"]}
+__doctest_requires__ = {("PerceptualEvaluationSpeechQuality"): ["pesq"]}
 
 
 class PerceptualEvaluationSpeechQuality(Metric):
@@ -134,35 +132,3 @@ def update(self, preds: Tensor, target: Tensor) -> None:  # type: ignore
     def compute(self) -> Tensor:
         """Computes average PESQ."""
         return self.sum_pesq / self.total
-
-
-class PESQ(PerceptualEvaluationSpeechQuality):
-    """Perceptual Evaluation of Speech Quality (PESQ).
-
-    .. deprecated:: v0.7
-        Use :class:`torchmetrics.audio.PerceptualEvaluationSpeechQuality`. Will be removed in v0.8.
-
-    Example:
-        >>> import torch
-        >>> g = torch.manual_seed(1)
-        >>> preds = torch.randn(8000)
-        >>> target = torch.randn(8000)
-        >>> nb_pesq = PESQ(8000, 'nb')
-        >>> nb_pesq(preds, target)
-        tensor(2.2076)
-        >>> wb_pesq = PESQ(16000, 'wb')
-        >>> wb_pesq(preds, target)
-        tensor(1.7359)
-    """
-
-    @deprecated(target=PerceptualEvaluationSpeechQuality, deprecated_in="0.7", remove_in="0.8", stream=_future_warning)
-    def __init__(
-        self,
-        fs: int,
-        mode: str,
-        compute_on_step: bool = True,
-        dist_sync_on_step: bool = False,
-        process_group: Optional[Any] = None,
-        dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
-    ) -> None:
-        void(fs, mode, compute_on_step, dist_sync_on_step, process_group, dist_sync_fn)

torchmetrics/audio/pit.py

@@ -13,12 +13,10 @@
 # limitations under the License.
 from typing import Any, Callable, Dict, Optional
 
-from deprecate import deprecated, void
 from torch import Tensor, tensor
 
 from torchmetrics.functional.audio.pit import permutation_invariant_training
 from torchmetrics.metric import Metric
-from torchmetrics.utilities import _future_warning
 
 
 class PermutationInvariantTraining(Metric):
@@ -58,11 +56,11 @@ class PermutationInvariantTraining(Metric):
     Example:
         >>> import torch
         >>> from torchmetrics import PermutationInvariantTraining
-        >>> from torchmetrics.functional import si_snr
+        >>> from torchmetrics.functional import scale_invariant_signal_noise_ratio
         >>> _ = torch.manual_seed(42)
         >>> preds = torch.randn(3, 2, 5) # [batch, spk, time]
         >>> target = torch.randn(3, 2, 5) # [batch, spk, time]
-        >>> pit = PermutationInvariantTraining(si_snr, 'max')
+        >>> pit = PermutationInvariantTraining(scale_invariant_signal_noise_ratio, 'max')
         >>> pit(preds, target)
         tensor(-2.1065)
 
@@ -114,39 +112,3 @@ def update(self, preds: Tensor, target: Tensor) -> None:  # type: ignore
     def compute(self) -> Tensor:
         """Computes average PermutationInvariantTraining metric."""
         return self.sum_pit_metric / self.total
-
-
-class PIT(PermutationInvariantTraining):
-    """Permutation invariant training (PIT). The PIT implements the famous Permutation Invariant Training method.
-
-    [1] in speech separation field in order to calculate audio metrics in a permutation invariant way.
-
-    .. deprecated:: v0.7
-        Use :class:`torchmetrics.audio.PermutationInvariantTraining`. Will be removed in v0.8.
-
-    Example:
-        >>> import torch
-        >>> from torchmetrics.functional import si_snr
-        >>> _ = torch.manual_seed(42)
-        >>> pit = PIT(si_snr, 'max')
-        >>> pit(torch.randn(3, 2, 5), torch.randn(3, 2, 5))
-        tensor(-2.1065)
-
-    Reference:
-        [1]	D. Yu, M. Kolbaek, Z.-H. Tan, J. Jensen, Permutation invariant training of deep models for
-        speaker-independent multi-talker speech separation, in: 2017 IEEE Int. Conf. Acoust. Speech
-        Signal Process. ICASSP, IEEE, New Orleans, LA, 2017: pp. 241–245. https://doi.org/10.1109/ICASSP.2017.7952154.
-    """
-
-    @deprecated(target=PermutationInvariantTraining, deprecated_in="0.7", remove_in="0.8", stream=_future_warning)
-    def __init__(
-        self,
-        metric_func: Callable,
-        eval_func: str = "max",
-        compute_on_step: bool = True,
-        dist_sync_on_step: bool = False,
-        process_group: Optional[Any] = None,
-        dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
-        **kwargs: Dict[str, Any],
-    ) -> None:
-        void(metric_func, eval_func, compute_on_step, dist_sync_on_step, process_group, dist_sync_fn, **kwargs)

torchmetrics/audio/sdr.py

@@ -13,15 +13,13 @@
 # limitations under the License.
 from typing import Any, Callable, Optional
 
-from deprecate import deprecated, void
 from torch import Tensor, tensor
 
 from torchmetrics.functional.audio.sdr import scale_invariant_signal_distortion_ratio, signal_distortion_ratio
 from torchmetrics.metric import Metric
-from torchmetrics.utilities import _future_warning
 from torchmetrics.utilities.imports import _FAST_BSS_EVAL_AVAILABLE
 
-__doctest_requires__ = {("SignalDistortionRatio", "SDR"): ["fast_bss_eval"]}
+__doctest_requires__ = {("SignalDistortionRatio"): ["fast_bss_eval"]}
 
 
 class SignalDistortionRatio(Metric):
@@ -156,50 +154,6 @@ def compute(self) -> Tensor:
         return self.sum_sdr / self.total
 
 
-class SDR(SignalDistortionRatio):
-    r"""Signal to Distortion Ratio (SDR)
-
-    .. deprecated:: v0.7
-        Use :class:`torchmetrics.SignalDistortionRatio`. Will be removed in v0.8.
-
-    Example:
-        >>> import torch
-        >>> g = torch.manual_seed(1)
-        >>> sdr = SDR()
-        >>> sdr(torch.randn(8000), torch.randn(8000))
-        tensor(-12.0589)
-        >>> # use with pit
-        >>> from torchmetrics.audio import PermutationInvariantTraining
-        >>> from torchmetrics.functional.audio import signal_distortion_ratio
-        >>> pit = PermutationInvariantTraining(signal_distortion_ratio, 'max')
-        >>> pit(torch.randn(4, 2, 8000), torch.randn(4, 2, 8000))
-        tensor(-11.6051)
-    """
-
-    @deprecated(target=SignalDistortionRatio, deprecated_in="0.7", remove_in="0.8", stream=_future_warning)
-    def __init__(
-        self,
-        use_cg_iter: Optional[int] = None,
-        filter_length: int = 512,
-        zero_mean: bool = False,
-        load_diag: Optional[float] = None,
-        compute_on_step: bool = True,
-        dist_sync_on_step: bool = False,
-        process_group: Optional[Any] = None,
-        dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
-    ) -> None:
-        void(
-            use_cg_iter,
-            filter_length,
-            zero_mean,
-            load_diag,
-            compute_on_step,
-            dist_sync_on_step,
-            process_group,
-            dist_sync_fn,
-        )
-
-
 class ScaleInvariantSignalDistortionRatio(Metric):
     """Scale-invariant signal-to-distortion ratio (SI-SDR). The SI-SDR value is in general considered an overall
     measure of how good a source sound.