|
from easydict import EasyDict |
|
|
|
walker_ddpg_config = dict( |
|
exp_name='evogym_walker_ddpg_seed0', |
|
env=dict( |
|
env_id='Walker-v0', |
|
robot='speed_bot', |
|
robot_dir='./dizoo/evogym/envs', |
|
collector_env_num=8, |
|
evaluator_env_num=8, |
|
n_evaluator_episode=8, |
|
stop_value=10, |
|
manager=dict(shared_memory=True, ), |
|
|
|
|
|
), |
|
policy=dict( |
|
cuda=True, |
|
|
|
random_collect_size=1000, |
|
model=dict( |
|
obs_shape=58, |
|
action_shape=10, |
|
twin_critic=False, |
|
actor_head_hidden_size=256, |
|
critic_head_hidden_size=256, |
|
action_space='regression', |
|
), |
|
learn=dict( |
|
update_per_collect=1, |
|
batch_size=256, |
|
learning_rate_actor=1e-3, |
|
learning_rate_critic=1e-3, |
|
ignore_done=False, |
|
target_theta=0.005, |
|
discount_factor=0.99, |
|
actor_update_freq=1, |
|
noise=False, |
|
), |
|
collect=dict( |
|
n_sample=1, |
|
unroll_len=1, |
|
noise_sigma=0.1, |
|
), |
|
other=dict(replay_buffer=dict(replay_buffer_size=1000000, ), ), |
|
) |
|
) |
|
walker_ddpg_config = EasyDict(walker_ddpg_config) |
|
main_config = walker_ddpg_config |
|
|
|
walker_ddpg_create_config = dict( |
|
env=dict( |
|
type='evogym', |
|
import_names=['dizoo.evogym.envs.evogym_env'], |
|
), |
|
env_manager=dict(type='subprocess'), |
|
policy=dict( |
|
type='ddpg', |
|
import_names=['ding.policy.ddpg'], |
|
), |
|
replay_buffer=dict(type='naive', ), |
|
) |
|
walker_ddpg_create_config = EasyDict(walker_ddpg_create_config) |
|
create_config = walker_ddpg_create_config |
|
|
|
if __name__ == "__main__": |
|
|
|
from ding.entry import serial_pipeline |
|
serial_pipeline((main_config, create_config), seed=0) |
|
|