fixup! Add test for dummy population axons

nengo_loihi/tests/test_conv.py

@@ -1075,16 +1075,26 @@ def test_conv_overlap_input(Simulator, plt):
 
 
 @pytest.mark.target_loihi
+@pytest.mark.parametrize("on_chip", [True, False])
 @pytest.mark.parametrize("precompute", [True, False])
 @pytest.mark.parametrize("pop_type", [16, 32])
 @pytest.mark.parametrize("channels_last", [True, False])
-def test_population_axons(precompute, pop_type, channels_last, Simulator, rng):
-    """On the chip, dummy axons were still having an effect. Check this is fixed."""
+def test_chip_population_axons(
+    on_chip, precompute, pop_type, channels_last, Simulator, rng
+):
+    """Check that all types of population axons work as inputs or between cores.
+
+    Also, on the chip, dummy axons were still having an effect. Check this is fixed.
+    """
 
-    def conv_layer(x, *args, label=None, **kwargs):
+    def conv_layer(input=None, *args, label=None, **kwargs):
         conv = nengo.Convolution(*args, **kwargs)
         layer = nengo.Ensemble(conv.output_shape.size, 1, label=label)
-        conn = nengo.Connection(x, layer.neurons, transform=conv)
+        conn = (
+            nengo.Connection(input, layer.neurons, transform=conv)
+            if input is not None
+            else None
+        )
         return layer, conv, conn
 
     if pop_type == 16 and not channels_last:
@@ -1102,7 +1112,7 @@ def conv_layer(x, *args, label=None, **kwargs):
     )
     X = rng.uniform(0.2, 1, size=input_shape.shape)
     kernel0 = rng.uniform(0.2, 1, size=(1, 1, 1, n_filters0))
-    kernel1 = rng.uniform(0.2, 1, size=(3, 3, n_filters0, n_filters1))
+    kernel1 = rng.uniform(0.1, 0.5, size=(3, 3, n_filters0, n_filters1))
 
     with nengo.Network(seed=0) as net:
         nengo_loihi.add_params(net)
@@ -1113,7 +1123,7 @@ def conv_layer(x, *args, label=None, **kwargs):
         net.config[nengo.Ensemble].intercepts = nengo.dists.Choice([0])
         net.config[nengo.Connection].synapse = 0.005
 
-        inp = nengo.Node(X.ravel())
+        inp = nengo.Node(X.ravel()) if not on_chip else None
 
         # first layer is off-chip to translate the inputs into spikes
         layer0, conv0, _ = conv_layer(
@@ -1125,7 +1135,14 @@ def conv_layer(x, *args, label=None, **kwargs):
             init=kernel0,
             label="layer0",
         )
-        net.config[layer0].on_chip = False
+
+        net.config[layer0].on_chip = on_chip
+        if on_chip:
+            assert kernel0.shape[:2] == (1, 1)
+            w = kernel0[0, 0]
+            Y = X.dot(w) if channels_last else np.tensordot(w.T, X, axes=1)
+            layer0.gain = nengo.dists.Choice([0.0])
+            layer0.bias = Y.ravel() * max_rate
 
         layer1, conv1, conn1 = conv_layer(
             layer0.neurons,