Working training - 00003 learning rate and frame skip

Run.py

@@ -2,6 +2,7 @@
 import gym
 from RandomAgent import TimeLimitWrapper
 from stable_baselines3 import PPO
+from stable_baselines3.common.atari_wrappers import MaxAndSkipEnv
 
 
 model = PPO.load("tmp/best_model.zip")
@@ -11,7 +12,8 @@ def main():
     #env = retro.make(game='MegaMan2-Nes')
     env = retro.make(game='SuperMarioBros-Nes')
     env = TimeLimitWrapper(env)
-
+    env = MaxAndSkipEnv(env, 4)
+
     obs = env.reset()
     done = False
 

Train.py

@@ -94,11 +94,11 @@ def _init():
     # You can choose between `DummyVecEnv` (usually faster) and `SubprocVecEnv`
     # env = make_vec_env(env_id, n_envs=num_cpu, seed=0, vec_env_cls=SubprocVecEnv)
 
-    #model = PPO('CnnPolicy', env, verbose=1, tensorboard_log="./board/")
-    model = PPO.load("tmp/best_model", env=env)
+    model = PPO('CnnPolicy', env, verbose=1, tensorboard_log="./board/", learning_rate=0.00003)
+    #model = PPO.load("tmp/best_model", env=env)
     print("------------- Start Learning -------------")
     callback = SaveOnBestTrainingRewardCallback(check_freq=1000, log_dir=log_dir)
-    model.learn(total_timesteps=5000000, callback=callback)
+    model.learn(total_timesteps=5000000, callback=callback, tb_log_name="PPO-00003")
     model.save(env_id)
     print("------------- Done Learning -------------")
     env = retro.make(game=env_id)