clean up fixture definitions of inference and summary networks

Author: paul-buerkner

tests/conftest.py

@@ -32,11 +32,9 @@ def conditions_size(request):
     return request.param
 
 
-@pytest.fixture(scope="function")
-def coupling_flow():
-    from bayesflow.networks import CouplingFlow
-
-    return CouplingFlow(depth=2, subnet="mlp", subnet_kwargs=dict(widths=(32, 32)))
+@pytest.fixture(params=[1, 4], scope="session")
+def summary_dim(request):
+    return request.param
 
 
 @pytest.fixture(params=["two_moons"], scope="session")
@@ -49,16 +47,6 @@ def feature_size(request):
     return request.param
 
 
-@pytest.fixture(params=["coupling_flow"], scope="function")
-def inference_network(request):
-    return request.getfixturevalue(request.param)
-
-
-@pytest.fixture(params=["inference_network", "summary_network"], scope="function")
-def network(request):
-    return request.getfixturevalue(request.param)
-
-
 @pytest.fixture(scope="session")
 def random_conditions(batch_size, conditions_size):
     if conditions_size is None:
@@ -94,13 +82,6 @@ def simulator(request):
     return request.getfixturevalue(request.param)
 
 
-@pytest.fixture(params=[None], scope="function")
-def summary_network(request, summary_dim):
-    if request.param is None:
-        return None
-    return request.getfixturevalue(request.param)
-
-
 @pytest.fixture(scope="session")
 def training_dataset(simulator, batch_size):
     from bayesflow.datasets import OfflineDataset

tests/test_networks/conftest.py

@@ -1,11 +1,6 @@
 import pytest
 
 
-@pytest.fixture(params=[1, 4], scope="session")
-def summary_dim(request):
-    return request.param
-
-
 # For the serialization tests, we want to test passing str and type.
 # For all other tests, this is not necessary and would double test time.
 # Therefore, below we specify two variants of each network, one without and

tests/test_two_moons/conftest.py

@@ -1,6 +1,13 @@
 import pytest
 
 
+@pytest.fixture()
+def inference_network():
+    from bayesflow.networks import CouplingFlow
+
+    return CouplingFlow(depth=2, subnet="mlp", subnet_kwargs=dict(widths=(32, 32)))
+
+
 @pytest.fixture()
 def approximator(adapter, inference_network):
     from bayesflow import ContinuousApproximator