def run_configuration(argv):
params = ParameterServer(
filename="examples/example_params/iqn_params.json")
params["ML"]["BaseAgent"][
"SummaryPath"] = "/home/mansoor/Study/Werkstudent/fortiss/code/bark-ml/summaries"
params["ML"]["BaseAgent"][
"CheckpointPath"] = "/home/mansoor/Study/Werkstudent/fortiss/code/bark-ml/checkpoints"
env = gym.make(FLAGS.env, params=params)
agent = IQNAgent(env=env, test_env=env, params=params)
if FLAGS.load and params["ML"]["BaseAgent"]["CheckpointPath"]:
agent.load_models(
os.path.join(params["ML"]["BaseAgent"]["CheckpointPath"], "best"))
if FLAGS.mode == "train":
agent.run()
elif FLAGS.mode == "visualize":
agent.visualize()
elif FLAGS.mode == "evaluate":
# writes evaluaion data using summary writer in summary path
agent.evaluate()
else:
raise Exception("Invalid argument for --mode")
behavior = BehaviorDiscreteMacroActionsML(params)
observer = NearestAgentsObserver(params)
evaluator = GoalReached(params)
viewer = MPViewer(
params=params,
center= [960, 1000.8],
enforce_x_length=True,
x_length = 100.0,
use_world_bounds=False)
# load env
env = HyDiscreteHighway(params=params,
scenario_generation=scenario_generator,
behavior=behavior,
evaluator=evaluator,
observer=observer,
viewer=viewer,
render=False)
# agent saved directory
agent_dir = os.path.join(exp_dir, 'agent')
# load agent
agent = IQNAgent(env=env, params=params, agent_save_dir=agent_dir,
is_checkpoint_run=True, is_online_demo=False)
agent.load_models(IQNAgent.check_point_directory(agent.agent_save_dir, "best"))
agent.evaluate()