训练后自动进行测试，自动删去不必要的文件

.gitignore

@@ -14,6 +14,7 @@ uv.lock
 *.pt
 *.pth
 runs/
+latest/
 
 # IDEs
 .vscode/
@@ -23,3 +24,6 @@ runs/
 # OS
 .DS_Store
 Thumbs.db
+
+# AI
+.claude/

main.py

@@ -2,6 +2,7 @@
 Main entry point for CTM-DQN speed limit control system.
 """
 import argparse
+import os
 from train import train
 from test import test
 
@@ -26,8 +27,8 @@ def main():
     parser.add_argument(
         "--model",
         type=str,
-        default=None,
-        help="Path to model checkpoint (for testing)",
+        default="latest/checkpoints/model_best.pt",
+        help="Path to model checkpoint (for testing, default: latest/checkpoints/model_best.pt)",
     )
 
     args = parser.parse_args()
@@ -37,7 +38,19 @@ def main():
         train(args.config)
     elif args.mode == "test":
         print("Starting testing mode...")
-        test(args.config, args.model)
+
+        # Auto-detect config from model path if using default model
+        config_path = args.config
+        model_path = args.model
+
+        # If using latest model and default config, use latest config instead
+        if model_path.startswith("latest/") and args.config == "config.yaml":
+            latest_config = os.path.join("latest", "config.yaml")
+            if os.path.exists(latest_config):
+                config_path = latest_config
+                print(f"Auto-detected config from latest run: {config_path}")
+
+        test(config_path, model_path)
 
 
 if __name__ == "__main__":

test.py

@@ -9,8 +9,14 @@ from environment import TrafficEnvironment
 from dqn_agent import DQNAgent
 
 
-def test(config_path: str = "config.yaml", model_path: str = None):
-    """Test trained DQN agent."""
+def test(config_path: str = "config.yaml", model_path: str = None, output_dir: str = None):
+    """Test trained DQN agent.
+
+    Args:
+        config_path: Path to configuration file
+        model_path: Path to model checkpoint
+        output_dir: Optional output directory for test results (overrides config)
+    """
     config = load_config(config_path)
     create_directories(config)
 
@@ -86,7 +92,9 @@ def test(config_path: str = "config.yaml", model_path: str = None):
     print("="*50)
 
     if render and len(all_densities) > 0:
-        plot_test_results(all_densities, all_actions, env, config["training"]["log_dir"])
+        # Use output_dir if specified, otherwise use config log_dir
+        log_dir = output_dir if output_dir else config["training"]["log_dir"]
+        plot_test_results(all_densities, all_actions, env, log_dir)
 
 
 def plot_test_results(densities, actions, env, log_dir):

train.py

@@ -11,6 +11,7 @@ from utils import load_config, create_run_directory
 from environment import TrafficEnvironment
 from parallel_env import ParallelEnvironment
 from dqn_agent import DQNAgent
+from test import test
 
 
 def set_random_seed(seed: int):
@@ -199,6 +200,13 @@ def train(config_path: str = "config.yaml"):
         episode_rewards, episode_losses, episode_throughputs, log_dir
     )
 
+    # Auto-test the best model
+    print("\n" + "="*60)
+    print("Starting automatic testing with best model...")
+    print("="*60)
+    config_copy_path = os.path.join(run_dir, "config.yaml")
+    test(config_path=config_copy_path, model_path=best_model_path, output_dir=log_dir)
+
 
 def plot_training_results(rewards, losses, throughputs, log_dir):
     """Plot and save training results."""

utils.py

@@ -31,6 +31,7 @@ def create_directories(config: Dict):
 def create_run_directory(config: Dict, config_path: str = "config.yaml") -> Tuple[str, str, str]:
     """
     Create a timestamped run directory for this training session.
+    Latest run is saved to 'latest/', previous runs are moved to 'runs/'.
 
     Args:
         config: Configuration dictionary
@@ -42,8 +43,37 @@ def create_run_directory(config: Dict, config_path: str = "config.yaml") -> Tupl
     # Create timestamp
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
 
-    # Create run directory
-    run_dir = os.path.join("runs", f"run_{timestamp}")
+    # Check if 'latest' directory exists
+    latest_dir = "latest"
+    if os.path.exists(latest_dir):
+        # Move existing 'latest' to 'runs' with timestamp from its config
+        try:
+            # Try to read timestamp from existing latest directory
+            old_config_path = os.path.join(latest_dir, "config.yaml")
+            if os.path.exists(old_config_path):
+                # Get modification time of the old config
+                mtime = os.path.getmtime(old_config_path)
+                old_timestamp = datetime.fromtimestamp(mtime).strftime("%Y%m%d_%H%M%S")
+            else:
+                old_timestamp = "unknown"
+
+            # Move to runs directory
+            os.makedirs("runs", exist_ok=True)
+            archive_dir = os.path.join("runs", f"run_{old_timestamp}")
+            shutil.move(latest_dir, archive_dir)
+
+            # Clean up timestamp files in archived directory
+            for file in os.listdir(archive_dir):
+                if file.startswith("timestamp_") and file.endswith(".txt"):
+                    timestamp_file_path = os.path.join(archive_dir, file)
+                    os.remove(timestamp_file_path)
+
+            print(f"Archived previous run to: {archive_dir}")
+        except Exception as e:
+            print(f"Warning: Could not archive previous run: {e}")
+
+    # Create new 'latest' directory
+    run_dir = latest_dir
     os.makedirs(run_dir, exist_ok=True)
 
     # Create subdirectories
@@ -52,11 +82,17 @@ def create_run_directory(config: Dict, config_path: str = "config.yaml") -> Tupl
     os.makedirs(checkpoint_dir, exist_ok=True)
     os.makedirs(log_dir, exist_ok=True)
 
-    # Copy config file to run directory
+    # Copy config file to run directory with timestamp
     config_copy_path = os.path.join(run_dir, "config.yaml")
     shutil.copy(config_path, config_copy_path)
 
+    # Also save a timestamped copy for reference
+    timestamp_file = os.path.join(run_dir, f"timestamp_{timestamp}.txt")
+    with open(timestamp_file, "w") as f:
+        f.write(f"Training started at: {timestamp}\n")
+
     print(f"Created run directory: {run_dir}")
+    print(f"Training timestamp: {timestamp}")
     print(f"Config saved to: {config_copy_path}")
 
     return run_dir, checkpoint_dir, log_dir