def demo2_ppo():
from ray_elegantrl.configs.configs_ppo import config
env = {
'id': 'AntBulletEnv-v0',
'state_dim': 28,
'action_dim': 8,
'if_discrete_action': False,
'reward_dim': 1,
'target_reward': 2500,
'max_step': 1000,
}
config['agent']['lambda_entropy'] = 0.05
config['agent']['lambda_gae_adv'] = 0.97
config['interactor']['rollout_num'] = 4
config['trainer']['sample_step'] = 2 ** 11 # env['max_step'] * config['interactor']['rollout_num']
config['trainer']['batch_size'] = 2 ** 11
config['interactor']['horizon_step'] = config['trainer']['sample_step']
config['trainer']['policy_reuse'] = 2 ** 3
config['interactor']['gamma'] = 0.99
config['evaluator']['break_step'] = int(8e6 * 8)
config['buffer']['max_buf'] = config['interactor']['horizon_step']
config['env'] = env
config['gpu_id'] = 2
config['if_cwd_time'] = True
config['random_seed'] = 0
beginer(config)
def demo1_ppo():
from ray_elegantrl.configs.configs_ppo import config
env = {
'id': 'ReacherBulletEnv-v0',
'state_dim': 9,
'action_dim': 2,
'if_discrete_action': False,
'reward_dim': 1,
'target_reward': 20,
'max_step': 150,
}
config['agent']['lambda_entropy'] = 0.01
config['agent']['lambda_gae_adv'] = 0.98
config['interactor']['rollout_num'] = 4
config['trainer'][
'sample_step'] = 4096 # env['max_step'] * config['interactor']['rollout_num']
config['interactor']['horizon_step'] = config['trainer']['sample_step']
config['trainer']['policy_reuse'] = 2**4
config['interactor']['gamma'] = 0.99
config['buffer']['max_buf'] = config['interactor']['horizon_step']
config['env'] = env
config['gpu_id'] = 2
config['if_cwd_time'] = True
config['random_seed'] = 10087
beginer(config)
def demo_carla_ppo():
from ray_elegantrl.configs.configs_ppo import config
from gym_carla_feature.env_a.mp_config import params
env = {
'id': 'carla-v0',
'state_dim': 42,
'action_dim': 2,
'if_discrete_action': False,
'reward_dim': 1,
'target_reward': 600,
'max_step': 512,
'params_name': {'params': params}
}
config['agent']['lambda_entropy'] = 0.05
config['agent']['lambda_gae_adv'] = 0.97
env['params_name']['params']['port'] = 2016
config['interactor']['rollout_num'] = 6
config['agent']['net_dim'] = 2 ** 9
config['trainer']['batch_size'] = 2 ** 9
config['trainer']['policy_reuse'] = 2 ** 3
config['interactor']['gamma'] = 0.99
config['trainer']['sample_step'] = 2 * env['max_step'] * config['interactor']['rollout_num']
config['interactor']['horizon_step'] = config['trainer']['sample_step']
config['env'] = env
config['gpu_id'] = 2
config['if_cwd_time'] = True
config['random_seed'] = 0
beginer(config)
def demo_carla_sac():
from ray_elegantrl.configs.configs_modsac import config
from gym_carla_feature.env_a.mp_config import params
env = {
'id': 'carla-v0',
'state_dim': 42,
'action_dim': 2,
'if_discrete_action': False,
'reward_dim': 1,
'target_reward': 600,
'max_step': 512,
'params_name': {
'params': params
}
}
config['interactor']['rollout_num'] = 4
config['trainer'][
'sample_step'] = env['max_step'] * config['interactor']['rollout_num']
config['interactor']['horizon_step'] = config['trainer']['sample_step']
config['interactor']['gamma'] = 0.99
config['env'] = env
config['gpu_id'] = 2
config['if_cwd_time'] = True
config['random_seed'] = 0
beginer(config)
def demo1_sac():
from ray_elegantrl.configs.configs_modsac import config
env = {
'id': 'ReacherBulletEnv-v0',
'state_dim': 9,
'action_dim': 2,
'if_discrete_action': False,
'reward_dim': 1,
'target_reward': 20,
'max_step': 150,
}
config['interactor']['rollout_num'] = 4
config['trainer']['sample_step'] = 1024 # env['max_step'] * config['interactor']['rollout_num']
config['interactor']['horizon_step'] = config['trainer']['sample_step']
config['interactor']['gamma'] = 0.99
config['env'] = env
config['gpu_id'] = 2
config['if_cwd_time'] = True
config['random_seed'] = 10087
beginer(config)
