Question How to train PPO algorithm in MLagents using Python instead of CLI?

I am seeking guidance on how to resolve the issue, or receive instructions on how to call the PPO algorithm using Python. As I need to configure several parameters in Python before executing PPO, the CLI training method is not suitable for my needs

I tried the following code, but it threw an error. I don't know if it's because the feature is not supported yet or if my code is incorrect. The error message is as follow

Code (Python):

1.  
2. from mlagents_envs.environment import UnityEnvironment
3. from mlagents.trainers.ppo.trainer import PPOTrainer
4. from mlagents.trainers.settings import TrainerSettings
5.  
6. trainer_settings = TrainerSettings()
7.  
8. env = UnityEnvironment("RollerAgent/RollerAgent.exe",no_graphics=True)
9. behavior_names = list(env.behavior_specs.keys())
10. print(behavior_names)
11. ppotrainer = PPOTrainer("RollerBall", 10, trainer_settings, True, False, 0, "")
12. env.reset()
13. for _ in range(1000):
14.     decision_steps, terminal_steps = env.get_steps(behavior_names[0])
15.     ppotrainer.advance()
16.  

error message
line 622, in _set_default_hyperparameters
return all_trainer_settings[self.trainer_type]()
KeyError: 'ppo'