|
from easydict import EasyDict |
|
|
|
agent_num = 8 |
|
collector_env_num = 16 |
|
evaluator_env_num = 8 |
|
|
|
main_config = dict( |
|
exp_name='smac_3s5z_coma_seed0', |
|
env=dict( |
|
map_name='3s5z', |
|
difficulty=7, |
|
reward_only_positive=True, |
|
mirror_opponent=False, |
|
agent_num=agent_num, |
|
collector_env_num=collector_env_num, |
|
evaluator_env_num=evaluator_env_num, |
|
stop_value=0.999, |
|
n_evaluator_episode=32, |
|
manager=dict( |
|
shared_memory=False, |
|
reset_timeout=6000, |
|
), |
|
), |
|
policy=dict( |
|
model=dict( |
|
|
|
|
|
agent_num=agent_num, |
|
|
|
|
|
|
|
|
|
obs_shape=dict( |
|
agent_state=150, |
|
global_state=216, |
|
), |
|
|
|
|
|
|
|
action_shape=14, |
|
|
|
actor_hidden_size_list=[64], |
|
), |
|
|
|
collect=dict( |
|
n_episode=32, |
|
env_num=collector_env_num, |
|
), |
|
eval=dict(env_num=evaluator_env_num, evaluator=dict(eval_freq=100, )), |
|
other=dict( |
|
eps=dict( |
|
type='exp', |
|
start=0.5, |
|
end=0.01, |
|
decay=200000, |
|
), |
|
replay_buffer=dict( |
|
|
|
replay_buffer_size=5000, |
|
|
|
max_use=10, |
|
), |
|
), |
|
), |
|
) |
|
main_config = EasyDict(main_config) |
|
create_config = dict( |
|
env=dict( |
|
type='smac', |
|
import_names=['dizoo.smac.envs.smac_env'], |
|
), |
|
env_manager=dict(type='subprocess'), |
|
policy=dict(type='coma'), |
|
collector=dict(type='episode', get_train_sample=True), |
|
) |
|
create_config = EasyDict(create_config) |
|
|
|
if __name__ == '__main__': |
|
|
|
from ding.entry import serial_pipeline |
|
serial_pipeline((main_config, create_config), seed=0) |
|
|