将每次运行参数及结果独立保存

.gitignore

@@ -11,10 +11,9 @@ wheels/
 uv.lock
 
 # Project specific - CTM-DQN
-checkpoints/
-logs/
 *.pt
 *.pth
+runs/
 
 # IDEs
 .vscode/

train.py

@@ -7,7 +7,7 @@ import numpy as np
 import torch
 from tqdm import tqdm
 import matplotlib.pyplot as plt
-from utils import load_config, create_directories
+from utils import load_config, create_run_directory
 from environment import TrafficEnvironment
 from parallel_env import ParallelEnvironment
 from dqn_agent import DQNAgent
@@ -28,7 +28,9 @@ def set_random_seed(seed: int):
 def train(config_path: str = "config.yaml"):
     """Train DQN agent."""
     config = load_config(config_path)
-    create_directories(config)
+
+    # Create timestamped run directory
+    run_dir, checkpoint_dir, log_dir = create_run_directory(config, config_path)
 
     # Set random seed for reproducibility
     random_seed = config["training"].get("random_seed", 42)
@@ -61,8 +63,8 @@ def train(config_path: str = "config.yaml"):
     num_episodes = config["training"]["num_episodes"]
     save_freq = config["training"]["save_freq"]
     log_freq = config["training"]["log_freq"]
-    checkpoint_dir = config["training"]["checkpoint_dir"]
     train_freq = config["training"].get("train_freq", 1)  # Train every N steps
+    # checkpoint_dir and log_dir are already set from create_run_directory
 
     episode_rewards = []
     episode_losses = []
@@ -194,7 +196,7 @@ def train(config_path: str = "config.yaml"):
     print(f"Best model saved to {best_model_path} (Best Reward: {best_reward:.2f})")
 
     plot_training_results(
-        episode_rewards, episode_losses, episode_throughputs, config["training"]["log_dir"]
+        episode_rewards, episode_losses, episode_throughputs, log_dir
     )
 
 

utils.py

@@ -3,7 +3,9 @@ Configuration utilities for loading and managing settings.
 """
 import yaml
 import os
-from typing import Dict
+import shutil
+from datetime import datetime
+from typing import Dict, Tuple
 
 
 def load_config(config_path: str = "config.yaml") -> Dict:
@@ -24,3 +26,37 @@ def create_directories(config: Dict):
 
     os.makedirs(checkpoint_dir, exist_ok=True)
     os.makedirs(log_dir, exist_ok=True)
+
+
+def create_run_directory(config: Dict, config_path: str = "config.yaml") -> Tuple[str, str, str]:
+    """
+    Create a timestamped run directory for this training session.
+
+    Args:
+        config: Configuration dictionary
+        config_path: Path to the config file
+
+    Returns:
+        Tuple of (run_dir, checkpoint_dir, log_dir)
+    """
+    # Create timestamp
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+
+    # Create run directory
+    run_dir = os.path.join("runs", f"run_{timestamp}")
+    os.makedirs(run_dir, exist_ok=True)
+
+    # Create subdirectories
+    checkpoint_dir = os.path.join(run_dir, "checkpoints")
+    log_dir = os.path.join(run_dir, "logs")
+    os.makedirs(checkpoint_dir, exist_ok=True)
+    os.makedirs(log_dir, exist_ok=True)
+
+    # Copy config file to run directory
+    config_copy_path = os.path.join(run_dir, "config.yaml")
+    shutil.copy(config_path, config_copy_path)
+
+    print(f"Created run directory: {run_dir}")
+    print(f"Config saved to: {config_copy_path}")
+
+    return run_dir, checkpoint_dir, log_dir