Improved PPO.

cair · Oct 30, 2024 · 11c944d · 11c944d
1 parent a280de8
commit 11c944d
Showing 1 changed file with 4 additions and 2 deletions.
diff --git a/jsp/examples/ppo_action.py b/jsp/examples/ppo_action.py
@@ -231,6 +231,8 @@ def get_best_schedule(self) -> Tuple[List[jobshop.Action], float]:
     def action_masks(self) -> np.ndarray:
         return self._action_mask
 
+    def get_jobshop_env(self) -> jobshop.JobShopEnvironment:
+        return self.env
 
 class MakespanCallback(BaseCallback):
     def __init__(self, verbose: int = 0, plotter: Optional[jobshop.LivePlotter] = None):
@@ -313,8 +315,8 @@ def make_env() -> Callable[[], gym.Env]:
     model.learn(total_timesteps=total_timesteps, callback=makespan_callback)
 
     print(f"Best makespan achieved: {makespan_callback.best_makespan}")
-    print(f"Optimal makespan: {ta_optimal}")
-    print(f"Gap: {(makespan_callback.best_makespan - ta_optimal) / ta_optimal * 100:.2f}%")
+    #print(f"Optimal makespan: {ta_optimal}")
+    #print(f"Gap: {(makespan_callback.best_makespan - ta_optimal) / ta_optimal * 100:.2f}%")
 
 if __name__ == "__main__":
     print(":D")