Added functionality to watch_agent.py to support evaluating different agents.

agents/watch_agent.py

@@ -5,8 +5,6 @@ import gymnasium as gym
 
 import argparse
 
-MODEL_NAME = "ALE-Pacman-v5"
-loaded_model = DQN.load(MODEL_NAME)
 
 # This script should have some options
 # 1. Turn off the stochasticity as determined by the ALEv5
@@ -18,15 +16,21 @@ loaded_model = DQN.load(MODEL_NAME)
 #       DONE
 # 4. Print the keyword args for the environment? I think this might be helpful...
 #       IN PROGRESS
-# 5. 
+# 5. Add option flag to accept file path for model
+# 6. Add option flag to accept number of episodes
 
 parser = argparse.ArgumentParser()
-parser.add_argument("-r", "--repeat_action_probability", help="repeat action probability", type=float, default=0.25)
-parser.add_argument("-f", "--frameskip", help="frameskip", type=int, default=4)
+parser.add_argument("-r", "--repeat_action_probability", help="repeat action probability, default 0.25", type=float, default=0.25)
+parser.add_argument("-f", "--frameskip", help="frameskip, default 4", type=int, default=4)
 parser.add_argument("-o", "--observe", help="observe agent", action="store_const", const=True)
 parser.add_argument("-p", "--print", help="print environment information", action="store_const", const=True)
+parser.add_argument("-e", "--num_episodes", help="specify the number of episodes to evaluate, default 1", type=int, default=1)
+parser.add_argument("-a", "--agent_filepath", help="file path to agent to watch, minus the .zip extension", type=str, required=True)
 args = parser.parse_args()
 
+MODEL_NAME = args.agent_filepath
+loaded_model = DQN.load(MODEL_NAME)
+
 # Toggle the render mode based on the -o flag
 if args.observe == True:
     mode = "human"
@@ -44,6 +48,7 @@ if args.print == True:
     for item in env_info:
         print(item)
 # Evaluate the policy
-mean_rwd, std_rwd = evaluate_policy(loaded_model.policy, eval_env, n_eval_episodes=1)
+mean_rwd, std_rwd = evaluate_policy(loaded_model.policy, eval_env, n_eval_episodes=args.num_episodes)
+print("eval episodes: ", args.num_episodes)
 print("mean rwd: ", mean_rwd)
 print("std rwd: ", std_rwd)