feat: Integrate task-specific graders into training script per hackathon rules

- Import TASK_GRADERS and get_grader_metadata from task_graders module
- Display available task graders with real-world applications at training startup
- Evaluate trained agent using task-specific grader function
- Log grader score on final test observation
- Ensures grader configuration is part of training pipeline per hackathon requirement

train_agent.py

@@ -1,6 +1,9 @@
 #!/usr/bin/env python3
 """
 Train an RL agent on the Energy Optimization Environment.
+
+Per hackathon requirements, this training script includes task-specific grader configuration
+to evaluate agent performance according to the defined scoring methodology.
 """
 
 import sys
@@ -19,14 +22,29 @@ sys.modules['he_demo'] = he_demo
 sys.modules['he_demo.models'] = he_demo
 
 from gym_wrapper import EnergyOptimizationGymEnv
+from task_graders import TASK_GRADERS, get_grader_metadata
 from stable_baselines3 import PPO
 from stable_baselines3.common.env_util import make_vec_env
 
 def train_agent():
-    """Train a PPO agent on the energy optimization environment."""
+    """Train a PPO agent on the energy optimization environment.
+    
+    Per hackathon requirements, this training configures task-specific graders
+    for evaluating agent performance according to defined scoring methodology.
+    """
 
     print("🚀 Training PPO Agent on Energy Optimization Environment")
     print("=" * 60)
+    
+    # ===== GRADER CONFIGURATION (Hackathon Requirement) =====
+    # Display available tasks and their grader configurations
+    print("\n📋 Available Task Graders:")
+    for task_name, task_info in TASK_GRADERS.items():
+        metadata = get_grader_metadata(task_name)
+        print(f"  • {metadata['display_name']} (Difficulty {metadata['difficulty']})")
+        print(f"    Targets: RAM < {metadata['target_ram']}%, Energy < {metadata['target_energy']} kWh")
+        print(f"    Application: {metadata['real_world_application']}")
+    print()
 
     # Create vectorized environment for better training
     def make_env():
@@ -60,19 +78,23 @@ def train_agent():
     print("✅ Model saved as 'energy_optimization_ppo.zip'")
 
     # Test the trained agent
-    print("\n🧪 Testing trained agent...")
+    print("\n🧪 Testing trained agent with grader evaluation...")
     test_env = EnergyOptimizationGymEnv()
     obs, _ = test_env.reset()
 
     total_reward = 0
     steps = 0
+    
+    # Import grader for evaluation
+    from task_graders import get_grader
+    grader_func = get_grader("balanced_optimization")  # Example grader task
 
     while steps < 50:
         # Get action from trained model
         action, _ = model.predict(obs, deterministic=True)
 
         # Execute action
-        obs, reward, done, _, _ = test_env.step(action)
+        obs, reward, done, _, info = test_env.step(action)
 
         total_reward += reward
         steps += 1
@@ -87,6 +109,16 @@ def train_agent():
 
         if done:
             break
+    
+    # Calculate grader score on final observation
+    if hasattr(test_env, 'env') and hasattr(test_env.env, 'observation'):
+        try:
+            final_obs = test_env.env.observation()
+            if final_obs and hasattr(final_obs, 'ram_usage'):
+                grader_score = grader_func(final_obs)
+                print(f"\n✅ Grader Score (Task: balanced_optimization): {grader_score:.3f}")
+        except Exception as e:
+            print(f"[DEBUG] Could not calculate grader score: {e}")
 
 if __name__ == "__main__":
     train_agent()