Add warning for 1-step training if synapse=None

Fixes #54

CHANGES.rst

@@ -21,6 +21,11 @@ Release History
 1.2.1 (unreleased)
 ------------------
 
+**Added**
+
+- Added a warning if users run one-timestep training with a network containing
+  synaptic filters.
+
 
 1.2.0 (September 5, 2018)
 -------------------------

nengo_dl/simulator.py

@@ -500,6 +500,14 @@ def train(self, inputs, targets, optimizer, n_epochs=1, objective="mse",
             raise ValidationError(
                 "Network was created with inference_only=True, cannot "
                 "be trained", "inference_only")
+        if (n_steps == 1 and self.model.toplevel is not None and
+                any(x.synapse is not None for x in
+                    (self.model.toplevel.all_connections +
+                     list(targets.keys())))):
+            warnings.warn(
+                "Training for one timestep, but the network contains "
+                "synaptic filters (which will introduce at least a "
+                "one-timestep delay); did you mean to set synapse=None?")
 
         # check for non-differentiable elements in graph
         # utils.find_non_differentiable(

nengo_dl/tests/test_simulator.py

@@ -1292,7 +1292,7 @@ def test_inference_only(Simulator, neuron_type, seed):
         # validation checks (can't do train/gradients in inference-only mode)
         with pytest.raises(ValidationError):
             sim2.train({a: np.zeros((1, 10, 1))}, {p: np.zeros((1, 10, 1))},
-                      tf.train.GradientDescentOptimizer(1))
+                       tf.train.GradientDescentOptimizer(1))
 
         with pytest.raises(ValidationError):
             sim2.check_gradients()
@@ -1302,3 +1302,41 @@ def test_dtype(Simulator):
     with pytest.warns(DeprecationWarning):
         with Simulator(None, dtype=tf.float32) as sim:
             assert sim.tensor_graph.dtype == tf.float32
+
+
+def test_synapse_warning(Simulator):
+    with nengo.Network() as net:
+        a = nengo.Node([0])
+        b = nengo.Ensemble(10, 1)
+        c = nengo.Connection(a, b, synapse=1)
+        p = nengo.Probe(b)
+        p2 = nengo.Probe(b)
+
+    # warning from connection
+    with Simulator(net) as sim:
+        with pytest.warns(UserWarning) as rec:
+            sim.train({a: np.zeros((1, 1, 1))}, {p: np.zeros((1, 1, 1))},
+                      tf.train.GradientDescentOptimizer(0))
+        assert any(str(w.message).startswith("Training for one timestep")
+                   for w in rec)
+
+    # warning from probe
+    c.synapse = None
+    p.synapse = 1
+    with Simulator(net) as sim:
+        with pytest.warns(UserWarning):
+            sim.train({a: np.zeros((1, 1, 1))}, {p: np.zeros((1, 1, 1))},
+                      tf.train.GradientDescentOptimizer(0))
+        assert any(str(w.message).startswith("Training for one timestep")
+                   for w in rec)
+
+    # no warning from non-target probe
+    p.synapse = None
+    p2.synapse = 1
+    with Simulator(net) as sim:
+        with pytest.warns(UserWarning) as rec:
+            sim.train({a: np.zeros((1, 1, 1))}, {p: np.zeros((1, 1, 1))},
+                      tf.train.GradientDescentOptimizer(0))
+            assert not any(
+                str(w.message).startswith("Training for one timestep")
+                for w in rec)