Trying to converge on simple columbus envs

test.py

@@ -2,91 +2,61 @@ import gym
 from gym.envs.registration import register
 import numpy as np
 import time
+import datetime
 
 from stable_baselines3 import SAC, PPO, A2C
 from stable_baselines3.common.evaluation import evaluate_policy
+from stable_baselines3.common.policies import ActorCriticCnnPolicy, ActorCriticPolicy, MultiInputActorCriticPolicy
 
 from sb3_trl.trl_pg import TRL_PG
-from columbus import env
+import columbus
 
-register(
-    id='ColumbusTestRay-v0',
-    entry_point=env.ColumbusTestRay,
-    max_episode_steps=30*60*5,
-)
 
-def main():
-    #env = gym.make("LunarLander-v2")
-    env = gym.make("ColumbusTestRay-v0")
-
-    ppo = PPO(
-        "MlpPolicy",
-        env,
-        verbose=1,
-        tensorboard_log="./logs_tb/test/ppo",
-        use_sde=False,
-        ent_coef=0.0001,
-        learning_rate=0.0004
-    )
-    ppo_base_sde = PPO(
-        "MlpPolicy",
-        env,
-        verbose=1,
-        tensorboard_log="./logs_tb/test/ppo_base_sde/",
-        use_sde=True,
-        sde_sample_freq=30*20,
-        sde_net_arch=[],
-        ent_coef=0.000001,
-        learning_rate=0.0003
-    )
+def main(env_name='ColumbusEasyObstacles-v0'):
+    env = gym.make(env_name)
     ppo_latent_sde = PPO(
         "MlpPolicy",
         env,
         verbose=1,
-        tensorboard_log="./logs_tb/test/ppo_latent_sde/",
+        tensorboard_log="./logs_tb/"+env_name+"/ppo_latent_sde/",
         use_sde=True,
-        sde_sample_freq=30*20,
-        ent_coef=0.000001,
-        learning_rate=0.0003
-    )
-    a2c = A2C(
-        "MlpPolicy",
-        env,
-        verbose=1,
-        tensorboard_log="./logs_tb/test/a2c/",
-    )
-    trl = TRL_PG(
-        "MlpPolicy",
-        env,
-        verbose=0,
-        tensorboard_log="./logs_tb/test/trl_pg/",
+        sde_sample_freq=30*15,
+        ent_coef=0.0032,
+        vf_coef=0.0005,
+        gamma=0.95,
+        learning_rate=0.02
     )
+    #trl = TRL_PG(
+    #    "MlpPolicy",
+    #    env,
+    #    verbose=0,
+    #    tensorboard_log="./logs_tb/"+env_name+"/trl_pg/",
+    #)
 
-    #print('PPO:')
-    #testModel(ppo, 500000, showRes = True, saveModel=True, n_eval_episodes=4)
-    print('PPO_BASE_SDE:')
-    testModel(ppo_base_sde, 200000, showRes = True, saveModel=True, n_eval_episodes=0)
-    #print('A2C:')
-    #testModel(a2c, showRes = True)
+    print('PPO_LATENT_SDE:')
+    testModel(ppo_latent_sde, 100000, showRes = True, saveModel=True, n_eval_episodes=0)
     #print('TRL_PG:')
     #testModel(trl)
 
 
-def testModel(model, timesteps=100000, showRes=False, saveModel=False, n_eval_episodes=16):
+def testModel(model, timesteps=150000, showRes=False, saveModel=False, n_eval_episodes=16):
     env = model.get_env()
     model.learn(timesteps)
 
+    if saveModel:
+        now = datetime.datetime.now().strftime('%d.%m.%Y-%H:%M')
+        model.save(model.tensorboard_log.replace('./logs_tb/','').replace('/','_')+now+'.zip')
+
     if n_eval_episodes:
         mean_reward, std_reward = evaluate_policy(model, env, n_eval_episodes=n_eval_episodes, deterministic=False)
         print('Reward: '+str(round(mean_reward,3))+'±'+str(round(std_reward,2)))
 
     if showRes:
-        model.save("model")
         input('<ready?>')
         obs = env.reset()
         # Evaluate the agent
         episode_reward = 0
-        for _ in range(1000):
+        for _ in range(30*60*5):
             time.sleep(1/30)
             action, _ = model.predict(obs, deterministic=False)
             obs, reward, done, info = env.step(action)