remote_base = os.getenv("RANGL_ENVIRONMENT_URL", "http://localhost:5000/")
client = Client(remote_base)
env_id = "reference-environment-v0"
seed = int(os.getenv("RANGL_SEED", 123456))
instance_id = client.env_create(env_id, seed)
client.env_monitor_start(
instance_id,
directory=f"monitor/{instance_id}",
force=True,
resume=False,
video_callable=False,
)
client.env_reset(instance_id)
while True:
action = client.env_action_space_sample(instance_id)
observation, reward, done, info = client.env_step(instance_id, action)
print(instance_id, reward)
if done:
print(instance_id)
break
client.env_monitor_close(instance_id)
print("done", done)
# make sure you print the instance_id as the last line in the script
print(instance_id)
env_id = "reference-environment-v0"
seed = int(os.getenv("RANGL_SEED", 123456))
instance_id = client.env_create(env_id, seed)
client.env_monitor_start(
instance_id,
directory=f"monitor/{instance_id}",
force=True,
resume=False,
video_callable=False,
)
model = DDPG.load("MODEL_ALPHA_GENERATION.zip")
observation = client.env_reset(instance_id)
print(observation)
import numpy as np
def ObservationTransform(obs, H, transform, steps_per_episode=int(96)):
step_count, generator_1_level, generator_2_level = obs[:3]
agent_prediction = np.array(obs[3:])
agent_horizon_prediction = agent_prediction[-1] * np.ones(
steps_per_episode)
agent_horizon_prediction[:int(steps_per_episode -
step_count)] = agent_prediction[int(
step_count):] # inclusive index