re-init evalCallback and save final model

Disabled eval logging to csv explicitely in the library, as this is
causing issues with the final progess*.csv

examples/shields/rl/13_minigridsb.py

@@ -89,21 +89,8 @@ def main():
         imageAndVideoCallback = ImageRecorderCallback(eval_env, render_freq, n_eval_episodes=1, evaluation_method=evaluate_policy, log_dir=log_dir, deterministic=True, verbose=0)
 
 
-    model.learn(steps,callback=[imageAndVideoCallback, InfoCallback()])
-
-    #vec_env = model.get_env()
-    #obs = vec_env.reset()
-    #terminated = truncated = False
-    #while not terminated and not truncated:
-    #    action_masks = None
-    #    action, _states = model.predict(obs, action_masks=action_masks)
-    #    print(action)
-    #    obs, reward, terminated, truncated, info = env.step(action)
-    #    # action, _states = model.predict(obs, deterministic=True)
-    #    # obs, rewards, dones, info = vec_env.step(action)
-    #    vec_env.render("human")
-    #    time.sleep(0.2)
-
+    model.learn(steps,callback=[imageAndVideoCallback, InfoCallback(), evalCallback])
+    model.save(f"{log_dir}/{expname(args)}")
 
 
 if __name__ == '__main__':