Fix a bug where BandPassFilter didn't work on GPU

tests/test_band_pass_filter.py

@@ -1,4 +1,5 @@
 import numpy as np
+import pytest
 import torch
 
 from torch_audiomentations import BandPassFilter
@@ -23,3 +24,25 @@ def test_band_pass_filter(self):
             ).samples.numpy()
             assert processed_samples.shape == samples.shape
             assert processed_samples.dtype == np.float32
+
+    @pytest.mark.skipif(not torch.cuda.is_available(), reason="Requires CUDA")
+    def test_band_pass_filter_cuda(self):
+        samples = np.array(
+            [
+                [[0.75, 0.5, -0.25, -0.125, 0.0], [0.65, 0.5, -0.25, -0.125, 0.0]],
+                [[0.3, 0.5, -0.25, -0.125, 0.0], [0.9, 0.5, -0.25, -0.125, 0.0]],
+                [[0.9, 0.5, -0.25, -1.06, 0.0], [0.9, 0.5, -0.25, -1.12, 0.0]],
+            ],
+            dtype=np.float32,
+        )
+        sample_rate = 16000
+
+        augment = BandPassFilter(p=1.0, output_type="dict")
+        for _ in range(20):
+            processed_samples = (
+                augment(samples=torch.from_numpy(samples).cuda(), sample_rate=sample_rate)
+                .samples.cpu()
+                .numpy()
+            )
+            assert processed_samples.shape == samples.shape
+            assert processed_samples.dtype == np.float32

torch_audiomentations/augmentations/band_pass_filter.py

@@ -112,7 +112,12 @@ def get_dist(min_freq, max_freq):
         )
 
         bandwidth_dist = torch.distributions.Uniform(
-            low=self.min_bandwidth_fraction, high=self.max_bandwidth_fraction,
+            low=torch.tensor(
+                self.min_bandwidth_fraction, dtype=torch.float32, device=samples.device
+            ),
+            high=torch.tensor(
+                self.max_bandwidth_fraction, dtype=torch.float32, device=samples.device
+            )
         )
         self.transform_parameters["bandwidth"] = bandwidth_dist.sample(
             sample_shape=(batch_size,)