[CI] Catch warnings we expect (asteroid-team#351)

julien-c · Nov 24, 2020 · c00403d · c00403d
1 parent cbd0bd0
commit c00403d
Show file tree

Hide file tree

Showing 8 changed files with 37 additions and 6 deletions.
diff --git a/asteroid/masknn/recurrent.py b/asteroid/masknn/recurrent.py
@@ -231,7 +231,7 @@ def forward(self, x):
         x = output.transpose(1, 2).transpose(2, -1).reshape(B * K, L, N)
         x = self.inter_RNN(x)
         x = self.inter_linear(x)
-        x = x.reshape(B, K, L, N).transpose(1, -1).transpose(2, -1)
+        x = x.reshape(B, K, L, N).transpose(1, -1).transpose(2, -1).contiguous()
         x = self.inter_norm(x)
         return output + x
 

diff --git a/pyproject.toml b/pyproject.toml
@@ -9,3 +9,8 @@ requires = [
 line-length = 100
 target-version = ["py36"]
 exclude = "(.eggs|.git|.hg|.mypy_cache|.nox|.tox|.venv|.svn|_build|buck-out|build|dist)"
+
+[tool.pytest.ini_options]
+filterwarnings = [
+    "ignore:Using or importing the ABCs.*:DeprecationWarning"
+]
diff --git a/tests/engine/__init__.py b/tests/engine/__init__.py
@@ -0,0 +1,5 @@
+import pytest
+
+ignored_warnings = ["ignore:Could not log computational graph since"]
+
+pytestmark = pytest.mark.filterwarnings(*ignored_warnings)
diff --git a/tests/filterbanks/transforms_test.py b/tests/filterbanks/transforms_test.py
@@ -67,6 +67,7 @@ def test_comp_mask(encoder_list):
         assert_allclose(masked, tf_rep)
 
 
+@pytest.mark.filterwarnings("ignore:asteroid.filterbanks.transforms.take_reim")
 def test_reim(encoder_list):
     for (enc, fb_dim) in encoder_list:
         tf_rep = enc(torch.randn(2, 1, 16000))  # [batch, freq, time]

diff --git a/tests/jit/__init__.py b/tests/jit/__init__.py
@@ -0,0 +1,10 @@
+import pytest
+
+ignored_warnings = [
+    "ignore:torch.tensor results are registered as constants in the trace.",
+    "ignore:Converting a tensor to a Python boolean might cause the trace to be incorrect.",
+    "ignore:Converting a tensor to a Python float might cause the trace to be incorrect.",
+    "ignore:Using or importing the ABCs from",
+]
+
+pytestmark = pytest.mark.filterwarnings(*ignored_warnings)
diff --git a/tests/losses/__init__.py b/tests/losses/__init__.py
@@ -0,0 +1,9 @@
+import pytest
+
+ignored_warnings = [
+    "ignore:Could not log computational graph since",
+    "ignore:The dataloader, val dataloader",
+    "ignore:The dataloader, train dataloader",
+]
+
+pytestmark = pytest.mark.filterwarnings(*ignored_warnings)
diff --git a/tests/models/demask_test.py b/tests/models/demask_test.py
@@ -63,7 +63,7 @@
 #         'activation': 'relu',
 #         'dropout': 0,
 #         'fb_kwargs': {},
-#         'fb_type': 'stft',
+#         'fb_name': 'stft',
 #         'hidden_dims': [1024],
 #         'input_type': 'mag',
 #         'kernel_size': 512,
@@ -78,7 +78,7 @@
 
 
 @pytest.mark.parametrize("input_type", ("mag", "cat", "reim"))
-@pytest.mark.parametrize("fb_type", ("stft", "free"))
+@pytest.mark.parametrize("fb_name", ("stft", "free"))
 @pytest.mark.parametrize("output_type", ("mag", "reim"))
 @pytest.mark.parametrize(
     "data",
@@ -90,11 +90,11 @@
         (torch.rand(2, 1, 50, requires_grad=False) - 0.5) * 2,
     ),
 )
-def test_forward(input_type, output_type, fb_type, data):
+def test_forward(input_type, output_type, fb_name, data):
     demask = DeMask(
         input_type=input_type,
         output_type=output_type,
-        fb_type=fb_type,
+        fb_name=fb_name,
         hidden_dims=(16,),
         kernel_size=8,
         n_filters=8,

diff --git a/tests/models/models_test.py b/tests/models/models_test.py
@@ -137,6 +137,7 @@ def test_sudormrf_imp():
     )
 
 
+@pytest.mark.filterwarnings("ignore: DPTransformer input dim")
 @pytest.mark.parametrize("fb", ["free", "stft", "analytic_free", "param_sinc"])
 def test_dptnet(fb):
     _default_test_model(DPTNet(2, ff_hid=10, chunk_size=4, n_repeats=2, fb_name=fb))
@@ -224,7 +225,7 @@ def test_available_models():
 
 @pytest.mark.parametrize("fb", ["free", "stft", "analytic_free", "param_sinc"])
 def test_demask(fb):
-    model = DeMask(fb_type=fb)
+    model = DeMask(fb_name=fb)
     test_input = torch.randn(1, 801)
 
     model_conf = model.serialize()