Apply deterministic RNG to more unit tests (#1173)

test/cut/test_custom_attrs.py

@@ -22,6 +22,7 @@
     dummy_recording,
     dummy_supervision,
 )
+from lhotse.testing.random import deterministic_rng
 
 
 @pytest.mark.parametrize("cut", [dummy_cut(1), dummy_cut(2).pad(300)])
@@ -103,7 +104,7 @@ def test_cut_custom_nonarray_attr_serialization():
     assert restored_cut.SNR == 7.3
 
 
-def test_cut_load_temporal_array():
+def test_cut_load_temporal_array(deterministic_rng):
     """Check that we can read a TemporalArray from a cut when their durations match."""
     alignment = np.random.randint(500, size=131)
     with TemporaryDirectory() as d, NumpyFilesWriter(d) as writer:
@@ -120,7 +121,7 @@ def test_cut_load_temporal_array():
         np.testing.assert_equal(alignment, restored_alignment)
 
 
-def test_cut_load_temporal_array_truncate():
+def test_cut_load_temporal_array_truncate(deterministic_rng):
     """Check the array loaded via TemporalArray is truncated along with the cut."""
     with TemporaryDirectory() as d, NumpyFilesWriter(d) as writer:
         expected_duration = 52.4  # 131 frames x 0.4s frame shift == 52.4s
@@ -138,7 +139,7 @@ def test_cut_load_temporal_array_truncate():
 
 
 @pytest.mark.parametrize("pad_value", [-1, 0])
-def test_cut_load_temporal_array_pad(pad_value):
+def test_cut_load_temporal_array_pad(deterministic_rng, pad_value):
     """Check the array loaded via TemporalArray is padded along with the cut."""
     with TemporaryDirectory() as d, NumpyFilesWriter(d) as writer:
         cut = MonoCut(
@@ -161,7 +162,7 @@ def test_cut_load_temporal_array_pad(pad_value):
         np.testing.assert_equal(alignment_pad[131:], pad_value)
 
 
-def test_validate_cut_with_temporal_array(caplog):
+def test_validate_cut_with_temporal_array(caplog, deterministic_rng):
     # Note: "caplog" is a special variable in pytest that captures logs.
     caplog.set_level(logging.WARNING)
     with TemporaryDirectory() as d, NumpyFilesWriter(d) as writer:
@@ -185,7 +186,7 @@ def test_validate_cut_with_temporal_array(caplog):
     )
 
 
-def test_cut_load_custom_recording():
+def test_cut_load_custom_recording(deterministic_rng):
     sampling_rate = 16000
     duration = 52.4
     audio = np.random.randn(1, compute_num_samples(duration, sampling_rate)).astype(
@@ -208,7 +209,7 @@ def test_cut_load_custom_recording():
         np.testing.assert_allclose(audio, restored_audio, atol=4e-5)
 
 
-def test_cut_load_custom_recording_truncate():
+def test_cut_load_custom_recording_truncate(deterministic_rng):
     sampling_rate = 16000
     duration = 52.4
     audio = np.random.randn(1, compute_num_samples(duration, sampling_rate)).astype(
@@ -235,7 +236,7 @@ def test_cut_load_custom_recording_truncate():
         np.testing.assert_allclose(audio[:, :80000], restored_audio, atol=3e-5)
 
 
-def test_cut_load_custom_recording_pad_right():
+def test_cut_load_custom_recording_pad_right(deterministic_rng):
     sampling_rate = 16000
     duration = 52.4
     audio = np.random.randn(1, compute_num_samples(duration, sampling_rate)).astype(
@@ -271,7 +272,7 @@ def test_cut_load_custom_recording_pad_right():
         np.testing.assert_allclose(0, restored_audio[:, audio.shape[1] :], atol=4e-5)
 
 
-def test_cut_load_custom_recording_pad_left():
+def test_cut_load_custom_recording_pad_left(deterministic_rng):
     sampling_rate = 16000
     duration = 52.4
     audio = np.random.randn(1, compute_num_samples(duration, sampling_rate)).astype(
@@ -307,7 +308,7 @@ def test_cut_load_custom_recording_pad_left():
         )
 
 
-def test_cut_load_custom_recording_pad_both():
+def test_cut_load_custom_recording_pad_both(deterministic_rng):
     sampling_rate = 16000
     duration = 52.4
     audio = np.random.randn(1, compute_num_samples(duration, sampling_rate)).astype(
@@ -350,7 +351,7 @@ def test_cut_load_custom_recording_pad_both():
         )
 
 
-def test_cut_attach_tensor():
+def test_cut_attach_tensor(deterministic_rng):
     alignment = np.random.randint(500, size=131)
     expected_duration = 52.4  # 131 frames x 0.4s frame shift == 52.4s
     cut = MonoCut(id="x", start=0, duration=expected_duration, channel=0)