Fix environment

src/scheduling/common/convert.py

@@ -73,7 +73,9 @@ def convert_to_proxy(
         n_shots=n_shots,
         noise=noise,
     )
-    preference = predict_device(tmp_circuit, [str(acc.uuid) for acc in accelerators])
+    preference = predict_device(
+        tmp_circuit, [str(acc.uuid) for acc in accelerators if acc is not None]
+    )
     if isinstance(circuit, QuantumCircuit):
         proxy.preselection = processing_time
         return proxy

src/scheduling/learning/environment.py

@@ -180,8 +180,8 @@ def reset(
             ],
             np.inf,
         )
-        for proxy in self.circuits:
-            assert isinstance(proxy, CircuitProxy)
+        for circuit in self.circuits:
+            proxy = convert_circuits([circuit], self.accelerators)[0]
             choice = next(
                 (
                     idx
@@ -216,7 +216,7 @@ def _get_info(self) -> dict[str, Any]:
                 for bucket in machine.buckets
             ]
             for machine in self._schedule.machines
-        } + {"schedule": self._schedule}
+        } | {"schedule": self._schedule}
 
     def _calculate_reward(self, completion_time: float, expected_noise: float) -> float:
         # Calculate the reward based on the completion time and expected noise