Reformat code with updated black

environment.yml

@@ -16,7 +16,7 @@ dependencies:
 - seaborn>=0.9,<1.0
 - pip:
   - audioread==2.1.8
-  - black
+  - black==23.12.1
   - coverage==5.3
   - julius>=0.2.3,<0.3
   - librosa==0.9.1

scripts/measure_convolve_execution_time.py

@@ -84,7 +84,7 @@
             execution_times[(t.description, num_examples)].append(t.execution_time)
 
     normalized_execution_times = {}
-    for (description, batch_size) in execution_times:
+    for description, batch_size in execution_times:
         times = execution_times[(description, batch_size)]
         if len(times) == 0:
             continue

tests/test_background_noise.py

@@ -69,7 +69,6 @@ def test_background_noise_no_guarantee_with_empty_tensor(self):
         self.assertEqual(mixed_input.size(0), self.empty_input_audio.size(0))
 
     def test_background_noise_guaranteed_with_zero_length_samples(self):
-
         with self.assertWarns(UserWarning) as warning_context_manager:
             mixed_input = self.bg_noise_transform_guaranteed(
                 self.empty_input_audio, self.sample_rate

tests/test_colored_noise.py

@@ -109,7 +109,9 @@ def test_invalid_params():
 
 def test_various_lengths_and_sample_rates():
     random.seed(42)
-    transform = AddColoredNoise(min_snr_in_db=10, max_snr_in_db=12, p=1.0, output_type="dict")
+    transform = AddColoredNoise(
+        min_snr_in_db=10, max_snr_in_db=12, p=1.0, output_type="dict"
+    )
 
     for _ in range(100):
         length = random.randint(1000, 100_000)

tests/test_gain.py

@@ -488,7 +488,6 @@ def test_reset_distribution(self):
         actual_gains_in_db = []
         for i in range(processed_samples.shape[0]):
             if not np.allclose(processed_samples[i], samples_batch[i]):
-
                 estimated_gain_factor = np.mean(processed_samples[i] / samples_batch[i])
                 estimated_gain_factor_in_db = convert_amplitude_ratio_to_decibels(
                     torch.tensor(estimated_gain_factor)
@@ -526,7 +525,6 @@ def test_cuda_reset_distribution(self):
         actual_gains_in_db = []
         for i in range(processed_samples.shape[0]):
             if not np.allclose(processed_samples[i], samples_batch[i]):
-
                 estimated_gain_factor = np.mean(processed_samples[i] / samples_batch[i])
                 estimated_gain_factor_in_db = convert_amplitude_ratio_to_decibels(
                     torch.tensor(estimated_gain_factor)

tests/test_impulse_response.py

@@ -108,7 +108,6 @@ def test_impulse_response_guaranteed_with_zero_length_samples(ir_transform):
 
 
 def test_impulse_response_access_file_paths(ir_path, sample_rate, input_audios):
-
     augment = ApplyImpulseResponse(
         ir_path, p=1.0, sample_rate=sample_rate, output_type="dict"
     )

tests/test_padding.py

@@ -9,7 +9,6 @@
 
 class TestPadding(unittest.TestCase):
     def test_padding_end(self):
-
         audio_samples = torch.rand(size=(2, 2, 32000), dtype=torch.float32)
         augment = Padding(
             min_fraction=0.2,
@@ -24,7 +23,6 @@ def test_padding_end(self):
         assert_almost_equal(padded_samples[..., -6400:].numpy(), np.zeros((2, 2, 6400)))
 
     def test_padding_start(self):
-
         audio_samples = torch.rand(size=(2, 2, 32000), dtype=torch.float32)
         augment = Padding(
             min_fraction=0.2,
@@ -39,7 +37,6 @@ def test_padding_start(self):
         assert_almost_equal(padded_samples[..., :6400].numpy(), np.zeros((2, 2, 6400)))
 
     def test_padding_zero(self):
-
         audio_samples = torch.rand(size=(2, 2, 32000), dtype=torch.float32)
         augment = Padding(min_fraction=0.2, max_fraction=0.5, p=0.0, output_type="dict")
         padded_samples = augment(audio_samples).samples
@@ -48,7 +45,6 @@ def test_padding_zero(self):
         assert_almost_equal(audio_samples.numpy(), padded_samples.numpy())
 
     def test_padding_perexample(self):
-
         audio_samples = torch.rand(size=(10, 2, 32000), dtype=torch.float32)
         augment = Padding(
             min_fraction=0.2,
@@ -72,7 +68,6 @@ def test_padding_perexample(self):
         self.assertLess(padded_samples.sum(), audio_samples.numpy().sum())
 
     def test_padding_perchannel(self):
-
         audio_samples = torch.rand(size=(10, 2, 32000), dtype=torch.float32)
         augment = Padding(
             min_fraction=0.2,
@@ -96,7 +91,6 @@ def test_padding_perchannel(self):
         self.assertLess(padded_samples.sum(), audio_samples.numpy().sum())
 
     def test_padding_variability_perexample(self):
-
         audio_samples = torch.rand(size=(10, 2, 32000), dtype=torch.float32)
         augment = Padding(
             min_fraction=0.2,
@@ -123,7 +117,6 @@ def test_padding_variability_perexample(self):
 
     @pytest.mark.skipif(not torch.cuda.is_available(), reason="Requires CUDA")
     def test_padding_cuda(self):
-
         audio_samples = torch.rand(
             size=(2, 2, 32000), dtype=torch.float32, device=torch.device("cuda")
         )

tests/test_random_crop.py

@@ -29,7 +29,6 @@ def test_crop_larger_cropto(self):
     @pytest.mark.skip(reason="output_type is not implemented yet")
     @pytest.mark.skipif(not torch.cuda.is_available(), reason="Requires CUDA")
     def test_crop_on_device_cuda(self):
-
         samples = torch.rand(
             size=(8, 2, 32000), dtype=torch.float32, device=torch.device("cuda")
         )

tests/test_spliceout.py

@@ -9,7 +9,6 @@
 
 class TestSpliceout(unittest.TestCase):
     def test_splice_out(self):
-
         audio_samples = torch.rand(size=(8, 1, 32000), dtype=torch.float32)
         augment = Compose(
             [
@@ -24,7 +23,6 @@ def test_splice_out(self):
         assert splice_out_samples.dtype == np.float32
 
     def test_splice_out_odd_hann(self):
-
         audio_samples = torch.rand(size=(8, 1, 32000), dtype=torch.float32)
         augment = Compose(
             [
@@ -39,7 +37,6 @@ def test_splice_out_odd_hann(self):
         assert splice_out_samples.dtype == np.float32
 
     def test_splice_out_per_batch(self):
-
         audio_samples = torch.rand(size=(8, 1, 32000), dtype=torch.float32)
         augment = Compose(
             [
@@ -62,7 +59,6 @@ def test_splice_out_per_batch(self):
         self.assertEqual(splice_out_samples.shape, audio_samples.shape)
 
     def test_splice_out_multichannel(self):
-
         audio_samples = torch.rand(size=(8, 2, 32000), dtype=torch.float32)
         augment = Compose(
             [
@@ -81,7 +77,6 @@ def test_splice_out_multichannel(self):
     @pytest.mark.skip(reason="This test fails and SpliceOut is not released yet")
     @pytest.mark.skipif(not torch.cuda.is_available(), reason="Requires CUDA")
     def test_splice_out_cuda(self):
-
         audio_samples = (
             torch.rand(
                 size=(8, 1, 32000), dtype=torch.float32, device=torch.device("cuda")

torch_audiomentations/augmentations/band_stop_filter.py

@@ -58,7 +58,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         perturbed = super().apply_transform(
             samples.clone(),
             sample_rate,

torch_audiomentations/augmentations/colored_noise.py

@@ -132,7 +132,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         batch_size, num_channels, num_samples = samples.shape
 
         # (batch_size, num_samples)

torch_audiomentations/augmentations/gain.py

@@ -81,7 +81,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         return ObjectDict(
             samples=samples * self.transform_parameters["gain_factors"],
             sample_rate=sample_rate,

torch_audiomentations/augmentations/high_pass_filter.py

@@ -49,7 +49,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         perturbed = super().apply_transform(
             samples=samples.clone(),
             sample_rate=sample_rate,

torch_audiomentations/augmentations/identity.py

@@ -24,7 +24,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         return ObjectDict(
             samples=samples,
             sample_rate=sample_rate,

torch_audiomentations/augmentations/impulse_response.py

@@ -83,7 +83,6 @@ def randomize_parameters(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ):
-
         batch_size, _, _ = samples.shape
 
         audio = self.audio if hasattr(self, "audio") else Audio(sample_rate, mono=True)
@@ -105,7 +104,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         batch_size, num_channels, num_samples = samples.shape
 
         # (batch_size, 1, max_ir_length)

torch_audiomentations/augmentations/low_pass_filter.py

@@ -111,7 +111,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         batch_size, num_channels, num_samples = samples.shape
 
         if self.cached_lpf is None:

torch_audiomentations/augmentations/mix.py

@@ -73,7 +73,6 @@ def randomize_parameters(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ):
-
         batch_size, num_channels, num_samples = samples.shape
         snr_distribution = torch.distributions.Uniform(
             low=torch.tensor(
@@ -109,7 +108,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         snr = self.transform_parameters["snr_in_db"]
         idx = self.transform_parameters["sample_idx"]
 

torch_audiomentations/augmentations/padding.py

@@ -7,7 +7,6 @@
 
 
 class Padding(BaseWaveformTransform):
-
     supported_modes = {"per_batch", "per_example", "per_channel"}
     supports_multichannel = True
     requires_sample_rate = False
@@ -70,7 +69,6 @@ def apply_transform(
         targets: Optional[int] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         for i, index in enumerate(self.transform_parameters["pad_length"]):
             if self.pad_section == "start":
                 samples[i, :, :index] = 0.0

torch_audiomentations/augmentations/peak_normalization.py

@@ -84,7 +84,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         if "divisors" in self.transform_parameters:
             samples[self.transform_parameters["selector"]] /= self.transform_parameters[
                 "divisors"

torch_audiomentations/augmentations/polarity_inversion.py

@@ -49,7 +49,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         return ObjectDict(
             samples=-samples,
             sample_rate=sample_rate,

torch_audiomentations/augmentations/random_crop.py

@@ -31,7 +31,6 @@ def __init__(
             raise ValueError('max_length_unit must be "samples" or "seconds"')
 
     def forward(self, samples, sampling_rate: typing.Optional[int] = None):
-
         sample_rate = sampling_rate or self.sampling_rate
         if sample_rate is None:
             raise RuntimeError("sample_rate is required")
@@ -74,7 +73,6 @@ def forward(self, samples, sampling_rate: typing.Optional[int] = None):
             (samples.shape[0], samples.shape[1], self.num_samples), device=samples.device
         )
         for i, sample in enumerate(samples):
-
             samples_cropped[i] = sample.unsqueeze(0)[
                 :, :, start_indices[i] : start_indices[i] + self.num_samples
             ]

torch_audiomentations/augmentations/shift.py

@@ -166,7 +166,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         num_samples_to_shift = self.transform_parameters["num_samples_to_shift"]
 
         # Select fastest implementation based on device

torch_audiomentations/augmentations/shuffle_channels.py

@@ -63,7 +63,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         if samples.shape[1] == 1:
             warnings.warn(
                 "Mono audio was passed to ShuffleChannels - there are no channels to shuffle."

torch_audiomentations/augmentations/splice_out.py

@@ -54,7 +54,6 @@ def randomize_parameters(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ):
-
         self.transform_parameters["splice_lengths"] = torch.randint(
             low=0,
             high=int(sample_rate * self.max_width * 1e-3),
@@ -68,11 +67,9 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         spliceout_samples = []
 
         for i in range(samples.shape[0]):
-
             random_lengths = self.transform_parameters["splice_lengths"][i]
             sample = samples[i][:, :]
             for j in range(self.num_time_intervals):

torch_audiomentations/augmentations/time_inversion.py

@@ -54,7 +54,6 @@ def apply_transform(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         # torch.flip() is supposed to be slower than np.flip()
         # An alternative is to use advanced indexing: https://github.com/pytorch/pytorch/issues/16424
         # reverse_index = torch.arange(selected_samples.size(-1) - 1, -1, -1).to(selected_samples.device)

torch_audiomentations/core/composition.py

@@ -102,7 +102,6 @@ def forward(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         inputs = ObjectDict(
             samples=samples,
             sample_rate=sample_rate,
@@ -191,7 +190,6 @@ def forward(
         targets: Optional[Tensor] = None,
         target_rate: Optional[int] = None,
     ) -> ObjectDict:
-
         inputs = ObjectDict(
             samples=samples,
             sample_rate=sample_rate,
@@ -200,7 +198,6 @@ def forward(
         )
 
         if random.random() < self.p:
-
             if not self.are_parameters_frozen:
                 self.randomize_parameters()