恢复训练时同步恢复历史记录以便绘图

train_sumo_appo_v2.py

@@ -100,6 +100,24 @@ def train_sumo_appo(resume_checkpoint=None):
     entropies = []
     best_reward = -float("inf")
 
+    # 加载历史训练数据
+    if resume_checkpoint:
+        import csv
+        log_file = os.path.join(log_dir, "appo_training_log.csv")
+        if os.path.exists(log_file):
+            with open(log_file, "r", encoding="utf-8") as f:
+                reader = csv.DictReader(f)
+                for row in reader:
+                    episode_rewards.append(float(row["reward"]))
+                    episode_throughputs.append(float(row["throughput"]))
+                    episode_mean_speeds.append(float(row["mean_speed"]))
+                    if row["policy_loss"]:
+                        policy_losses.append(float(row["policy_loss"]))
+                    if row["value_loss"]:
+                        value_losses.append(float(row["value_loss"]))
+            best_reward = max(episode_rewards) if episode_rewards else -float("inf")
+            print(f"已加载 {len(episode_rewards)} 条历史记录")
+
     print("开始训练...\n")
 
     try:

train_sumo_dqn.py

@@ -95,6 +95,22 @@ def train_sumo_dqn(resume_checkpoint=None):
     losses = []
     best_reward = -float("inf")
 
+    # 加载历史训练数据
+    if resume_checkpoint:
+        import csv
+        log_file = os.path.join(log_dir, "dqn_training_log.csv")
+        if os.path.exists(log_file):
+            with open(log_file, "r", encoding="utf-8") as f:
+                reader = csv.DictReader(f)
+                for row in reader:
+                    episode_rewards.append(float(row["reward"]))
+                    episode_throughputs.append(float(row["throughput"]))
+                    episode_mean_speeds.append(float(row["mean_speed"]))
+                    if row["value_loss"]:
+                        losses.append(float(row["value_loss"]))
+            best_reward = max(episode_rewards) if episode_rewards else -float("inf")
+            print(f"已加载 {len(episode_rewards)} 条历史记录")
+
     print("开始训练...\n")
 
     try:

train_sumo_ppo.py

@@ -104,6 +104,24 @@ def train_sumo_ppo(resume_checkpoint=None):
     entropies = []
     best_reward = -float("inf")
 
+    # 加载历史训练数据
+    if resume_checkpoint:
+        import csv
+        log_file = os.path.join(log_dir, "ppo_training_log.csv")
+        if os.path.exists(log_file):
+            with open(log_file, "r", encoding="utf-8") as f:
+                reader = csv.DictReader(f)
+                for row in reader:
+                    episode_rewards.append(float(row["reward"]))
+                    episode_throughputs.append(float(row["throughput"]))
+                    episode_mean_speeds.append(float(row["mean_speed"]))
+                    if row["policy_loss"]:
+                        policy_losses.append(float(row["policy_loss"]))
+                    if row["value_loss"]:
+                        value_losses.append(float(row["value_loss"]))
+            best_reward = max(episode_rewards) if episode_rewards else -float("inf")
+            print(f"已加载 {len(episode_rewards)} 条历史记录")
+
     print("开始训练...\n")
 
     try: