cost correction

montrealrobotics · Nov 16, 2023 · 7dc92bb · 7dc92bb
1 parent 787f088
commit 7dc92bb
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/cleanrl/ddpg_continuous_action.py b/cleanrl/ddpg_continuous_action.py
@@ -462,7 +462,7 @@ def convert_dict_to_tensor(data, device):
         for i, info in enumerate(infos):
             if 'episode' in info.keys():
                 ep_len = info["episode"]["l"]
-                total_cost += infos[0]["cum_cost"]
+                total_cost += infos[0]["cost"]
                 # num_successes += int(infos[0]["cum_goal_met"])
                 success_rate.append(int(infos[0]["is_success"]))
                 score.append(info["episode"]['r'])
@@ -475,7 +475,7 @@ def convert_dict_to_tensor(data, device):
                 writer.add_scalar("charts/Total Success", num_successes, global_step)
                 writer.add_scalar("charts/Success rate", np.mean(success_rate[-100:]), global_step)
                 writer.add_scalar("charts/Avg. Return", np.mean(score[-100:]), global_step)
-                writer.add_scalar("cost/ep_cost", info["cum_cost"], global_step)
+                writer.add_scalar("cost/ep_cost", info["cost"], global_step)
                 writer.add_scalar("cost/total_cost", total_cost, global_step)
 
                 e_risks = torch.Tensor(np.array(list(reversed(range(int(ep_len))))) if info["cost"] > 0 else np.array([int(ep_len)]*int(ep_len))).repeat_interleave(2).numpy()