Reject quadratic biases for unknown edges

dwave/plugins/torch/models/boltzmann_machine.py

@@ -134,6 +134,8 @@ def set_quadratic(self, quadratic: dict[tuple[Hashable, Hashable], float]) -> No
         """
         for (u, v), bias in quadratic.items():
             idx = self._edge_to_idx.get((u, v), self._edge_to_idx.get((v, u)))
+            if idx is None:
+                raise ValueError(f"Edge {(u, v)!r} is not in the model.")
             self._quadratic.data[idx] = bias
 
     def _setup_hidden(self):

releasenotes/notes/reject-unknown-quadratic-edges-d32fa47da9b7380f.yaml

@@ -0,0 +1,5 @@
+---
+fixes:
+  - |
+    Raise a ``ValueError`` when ``GraphRestrictedBoltzmannMachine.set_quadratic`` receives an edge
+    that is not in the model. Previously, an unknown edge overwrote every quadratic bias.

tests/test_boltzmann_machine.py

@@ -82,10 +82,18 @@ def test_selfloop(self):
             GRBM(self.nodes, self.edges, None, {"a": 0}, {("b", "c"): 0})
 
     def test_quadratic(self):
-        self.bm.set_quadratic({("d", "b"): 999})
+        self.bm.set_quadratic({("b", "a"): 999})
         self.assertEqual(999, self.bm.quadratic[0])
         self.bm.set_quadratic({})
 
+    def test_set_quadratic_unknown_edge(self):
+        quadratic = self.bm.quadratic.detach().clone()
+
+        with self.assertRaisesRegex(ValueError, r"Edge \('d', 'b'\) is not in the model"):
+            self.bm.set_quadratic({("d", "b"): 999})
+
+        torch.testing.assert_close(self.bm.quadratic, quadratic)
+
     def test_set_linear(self):
         self.bm.set_linear({"d": 999})
         self.assertEqual(999, self.bm.linear[0])