|
|
|
|
|
|
|
|
|
|
|
def test_trajectory_to_agentbuffer(): |
|
|
|
length = 15 |
|
|
|
# These keys should be of type np.ndarray |
|
|
|
wanted_keys = [ |
|
|
|
(ObservationKeyPrefix.OBSERVATION, 0), |
|
|
|
(ObservationKeyPrefix.OBSERVATION, 1), |
|
|
|
|
|
|
|
BufferKey.ACTION_MASK, |
|
|
|
BufferKey.PREV_ACTION, |
|
|
|
BufferKey.ENVIRONMENT_REWARDS, |
|
|
|
BufferKey.GROUP_REWARD, |
|
|
|
# These keys should be of type List |
|
|
|
"group_obs_0", |
|
|
|
"group_obs_1", |
|
|
|
"group_obs_next_0", |
|
|
|
"group_obs_next_1", |
|
|
|
"groupmate_rewards", |
|
|
|
"group_dones", |
|
|
|
BufferKey.GROUPMATE_REWARDS, |
|
|
|
BufferKey.GROUP_CONTINUOUS_ACTION, |
|
|
|
BufferKey.GROUP_DISCRETE_ACTION, |
|
|
|
BufferKey.GROUP_DONES, |
|
|
|
BufferKey.GROUP_NEXT_CONT_ACTION, |
|
|
|
BufferKey.GROUP_NEXT_DISC_ACTION, |
|
|
|
] |
|
|
|
wanted_keys = set(wanted_keys + wanted_group_keys) |
|
|
|
trajectory = make_fake_trajectory( |
|
|
|
|
Ervin Teng