Fix headless argument behaviour in ULTRA (#737)

* Fix headless argument in ULTRA scripts

* Change headless flag to behave like SMARTS

ultra/docs/getting_started.md

@@ -52,7 +52,7 @@ Implementations of baseline agents are available in `ultra/baselines/`. Notice,
   - `--level`: The level of the task (default is easy).
   - `--episodes`: The number of training episodes to run (default is 1000000).
   - `--timestep`: The environment timestep in seconds (default is 0.1).
-  - `--headless`: Whether to run training without Envision (default is True).
+  - `--headless`: Provide this flag to run training without Envision.
   - `--eval-episodes`: The number of evaluation episodes (default is 200).
   - `--eval-rate`: The rate at which evaluation occurs based on the number of observations (default is 10000).
   - `--seed`: The environment seed (default is 2).
@@ -82,7 +82,7 @@ After training your agent, your models should be saved under `logs/<timestamped_
   - `--models`: The path to the saved model (default is models/).
   - `--episodes`: The number of evaluation episodes (default is 200).
   - `--timestep`: The environment timestep in seconds (default is 0.1).
-  - `--headless`: Whether to run evaluation without Envision (default is True).
+  - `--headless`: Provide this flag to run evaluation without Envision.
   - `--experiment-dir`: The path to the spec file that includes adapters and policy parameters.
   - `--policy`: The policy (agent) to evaluate (default is sac).
   - `--max-steps-episode`: The option to limit the number of steps per epsiodes (default is 10000).

ultra/docs/rllib.md

@@ -26,7 +26,7 @@ of the steps is shown below
   - `--level`: The level of the task (default is easy).
   - `--episodes`: The number of training episodes to run (default is 100).
   - `--timestep`: The environment timestep in seconds (default is 0.1).
-  - `--headless`: Whether to run training without Envision (default is True).
+  - `--headless`: Provide this flag to run training without Envision.
   - `--eval-episodes`: The number of evaluation episodes (default is 200).
   - `--eval-rate`: The rate at which evaluation occurs based on the number of episodes (default is 10000).
   - `--seed`: The environment seed (default is 2).

ultra/tests/test_evaluate.py

@@ -63,12 +63,12 @@ def setUpClass(cls):
         )
         train_command = (
             "python ultra/train.py "
-            "--task 00 --level eval_test --policy sac --headless True --episodes 1 "
+            "--task 00 --level eval_test --policy sac --headless --episodes 1 "
             f"--eval-rate 1 --eval-episodes 1 --max-episode-steps 2 --log-dir {path}"
         )
         multiagent_train_command = (
             "python ultra/train.py "
-            "--task 00-multiagent --level eval_test --policy sac,dqn,ppo --headless True --episodes 1 "
+            "--task 00-multiagent --level eval_test --policy sac,dqn,ppo --headless --episodes 1 "
             f"--eval-rate 1 --eval-episodes 1 --max-episode-steps 2 --log-dir {multiagent_path}"
         )
 
@@ -174,7 +174,7 @@ def test_evaluate_cli(self):
         evaluate_command = (
             f"python ultra/evaluate.py "
             f"--task 00 --level eval_test --models {models} --experiment-dir {experiment_dir} "
-            f"--episodes 1 --max-episode-steps 2 --log-dir {log_dir} --headless True"
+            f"--episodes 1 --max-episode-steps 2 --log-dir {log_dir} --headless"
         )
 
         ray.shutdown()
@@ -202,7 +202,7 @@ def test_evaluate_cli_multiagent(self):
         evaluate_command = (
             f"python ultra/evaluate.py "
             f"--task 00-multiagent --level eval_test --models {models} --experiment-dir {experiment_dir} "
-            f"--episodes 1 --max-episode-steps 2 --log-dir {log_dir} --headless True"
+            f"--episodes 1 --max-episode-steps 2 --log-dir {log_dir} --headless"
         )
 
         ray.shutdown()

ultra/tests/test_rllib_train.py

@@ -42,7 +42,7 @@ def test_rllib_train_cli(self):
         log_dir = os.path.join(RLlibTrainTest.OUTPUT_DIRECTORY, "tests/rllib_results/")
         try:
             os.system(
-                f"python ultra/rllib_train.py --task 00 --level easy --episodes 1 --training-batch-size 200 --headless True --log-dir {log_dir}"
+                f"python ultra/rllib_train.py --task 00 --level easy --episodes 1 --training-batch-size 200 --headless --log-dir {log_dir}"
             )
         except Exception as err:
             print(err)

ultra/ultra/evaluate.py

@@ -226,7 +226,9 @@ def evaluate(
         "--timestep", help="Environment timestep (sec)", type=float, default=0.1
     )
     parser.add_argument(
-        "--headless", help="Run without envision", type=bool, default=True
+        "--headless",
+        help="Run without envision",
+        action="store_true",
     )
     parser.add_argument(
         "--experiment-dir",

ultra/ultra/rllib_train.py

@@ -179,7 +179,9 @@ def train(
         "--timestep", help="environment timestep (sec)", type=float, default=0.1
     )
     parser.add_argument(
-        "--headless", help="run without envision", type=bool, default=True
+        "--headless",
+        help="Run without envision",
+        action="store_true",
     )
     parser.add_argument(
         "--eval-episodes", help="number of evaluation episodes", type=int, default=100

ultra/ultra/train.py

@@ -224,7 +224,9 @@ def train(
         "--timestep", help="Environment timestep (sec)", type=float, default=0.1
     )
     parser.add_argument(
-        "--headless", help="Run without envision", type=bool, default=True
+        "--headless",
+        help="Run without envision",
+        action="store_true",
     )
     parser.add_argument(
         "--eval-episodes", help="Number of evaluation episodes", type=int, default=200