def main():
register_env("pa_cartpole", lambda _: ParametricActionsCartPole(10))
trainer = RandomParametricTrainer(env="pa_cartpole")
result = trainer.train()
assert result["episode_reward_mean"] > 10, result
print("Test: OK")
from ray.rllib.utils.test_utils import check_learning_achieved
from ray.tune.registry import register_env
parser = argparse.ArgumentParser()
parser.add_argument("--run", type=str, default="PPO")
parser.add_argument("--torch", action="store_true")
parser.add_argument("--as-test", action="store_true")
parser.add_argument("--stop-iters", type=int, default=200)
parser.add_argument("--stop-reward", type=float, default=150.0)
parser.add_argument("--stop-timesteps", type=int, default=100000)
if __name__ == "__main__":
args = parser.parse_args()
ray.init()
register_env("pa_cartpole", lambda _: ParametricActionsCartPole(10))
ModelCatalog.register_custom_model(
"pa_model",
TorchParametricActionsModel if args.torch else ParametricActionsModel)
if args.run == "DQN":
cfg = {
# TODO(ekl) we need to set these to prevent the masked values
# from being further processed in DistributionalQModel, which
# would mess up the masking. It is possible to support these if we
# defined a custom DistributionalQModel that is aware of masking.
"hiddens": [],
"dueling": False,
}
else:
cfg = {}
from ray.rllib.models import ModelCatalog
from ray.rllib.utils.test_utils import check_learning_achieved
from ray.tune.registry import register_env
import ray.rllib.agents.ppo as ppo
parser = argparse.ArgumentParser()
parser.add_argument("--run", type=str, default="PPO")
parser.add_argument("--stop-iters", type=int, default=200)
parser.add_argument("--stop-reward", type=float, default=150.0)
parser.add_argument("--stop-timesteps", type=int, default=100000)
if __name__ == "__main__":
args = parser.parse_args()
ray.init()
register_env("pa_cartpole", lambda _: ParametricActionsCartPole(10))
ModelCatalog.register_custom_model("pa_model", TorchParametricActionsModel)
if args.run == "DQN":
cfg = {
# TODO(ekl) we need to set these to prevent the masked values
# from being further processed in DistributionalQModel, which
# would mess up the masking. It is possible to support these if we
# defined a custom DistributionalQModel that is aware of masking.
"hiddens": [],
"dueling": False,
}
else:
cfg = {}
config = dict(