openrl.modules package

Subpackages

• openrl.modules.common package
  • Submodules
  • openrl.modules.common.a2c_net module
  • openrl.modules.common.base_net module
  • openrl.modules.common.bc_net module
  • openrl.modules.common.ddpg_net module
  • openrl.modules.common.dqn_net module
  • openrl.modules.common.gail_net module
  • openrl.modules.common.mat_net module
  • openrl.modules.common.ppo_net module
  • openrl.modules.common.sac_net module
  • openrl.modules.common.vdn_net module
  • Module contents
• openrl.modules.networks package
  • Subpackages
    • openrl.modules.networks.utils package
      • Subpackages
      • Submodules
      • openrl.modules.networks.utils.act module
      • openrl.modules.networks.utils.attention module
      • openrl.modules.networks.utils.cnn module
      • openrl.modules.networks.utils.distributed_utils module
      • openrl.modules.networks.utils.distributions module
      • openrl.modules.networks.utils.mix module
      • openrl.modules.networks.utils.mlp module
      • openrl.modules.networks.utils.popart module
      • openrl.modules.networks.utils.rnn module
      • openrl.modules.networks.utils.running_mean_std module
      • openrl.modules.networks.utils.transformer_act module
      • openrl.modules.networks.utils.util module
      • openrl.modules.networks.utils.vdn module
      • Module contents
  • Submodules
  • openrl.modules.networks.MAT_network module
  • openrl.modules.networks.base_policy_network module
  • openrl.modules.networks.base_value_network module
  • openrl.modules.networks.base_value_policy_network module
  • openrl.modules.networks.ddpg_network module
  • openrl.modules.networks.gail_discriminator module
  • openrl.modules.networks.policy_network module
  • openrl.modules.networks.policy_network_gpt module
  • openrl.modules.networks.policy_value_network module
  • openrl.modules.networks.policy_value_network_gpt module
  • openrl.modules.networks.policy_value_network_sb3 module
  • openrl.modules.networks.q_network module
  • openrl.modules.networks.sac_network module
  • openrl.modules.networks.value_network module
  • openrl.modules.networks.value_network_gpt module
  • openrl.modules.networks.vdn_network module
  • Module contents
• openrl.modules.utils package
  • Submodules
  • openrl.modules.utils.util module
  • openrl.modules.utils.valuenorm module
  • Module contents

Submodules

openrl.modules.base_module module

class openrl.modules.base_module.BaseModule(cfg)

Bases: abc.ABC

convert_distributed_model() → None

abstract lr_decay(episode: int, episodes: int) → None

abstract restore(model_dir: str) → None

abstract save(save_dir: str) → None

openrl.modules.bc_module module

class openrl.modules.bc_module.BCModule(cfg, policy_input_space: gym.spaces.box.Box, critic_input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, share_model: bool = False, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, rnn_states_actor, masks, action_masks=None, deterministic=False)

evaluate_actions(critic_obs, obs, rnn_states_actor, rnn_states_critic, action, masks, action_masks=None, active_masks=None, critic_masks_batch=None)

get_actions(critic_obs, obs, rnn_states_actor, rnn_states_critic, masks, action_masks=None, deterministic=False)

get_critic_value_normalizer()

get_model_configs(cfg) → Dict[str, Any]

get_values(critic_obs, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

openrl.modules.ddpg_module module

class openrl.modules.ddpg_module.DDPGModule(cfg, input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, deterministic: bool)

evaluate_actions(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, action_masks=None, masks_batch=None)

evaluate_actor_loss(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, action_masks=None, masks_batch=None)

evaluate_critic_loss(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, next_masks_batch, action_masks=None, masks_batch=None)

get_actions(obs)

get_critic_value_normalizer()

get_values(obs, action, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

openrl.modules.dqn_module module

class openrl.modules.dqn_module.DQNModule(cfg, input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, rnn_states_actor, masks, action_masks=None)

evaluate_actions(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, next_masks, action_masks=None, masks_batch=None, critic_masks_batch=None)

get_actions(obs, rnn_states, masks, action_masks=None)

get_critic_value_normalizer()

get_values(obs, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

openrl.modules.gail_module module

class openrl.modules.gail_module.GAILModule(cfg, policy_input_space: gym.spaces.box.Box, critic_input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, share_model: bool = False, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.ppo_module.PPOModule

get_model_configs(cfg) → Dict[str, Any]

openrl.modules.model_config module

class openrl.modules.model_config.ModelConfig(*args, **kwargs)

Bases: dict

class openrl.modules.model_config.ModelTrainConfig(model: torch.nn.modules.module.Module, input_space: gym.spaces.box.Box, lr: Optional[float] = None, *args, **kwargs)

Bases: openrl.modules.model_config.ModelConfig

openrl.modules.ppo_module module

class openrl.modules.ppo_module.PPOModule(cfg, policy_input_space: gym.spaces.box.Box, critic_input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, share_model: bool = False, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, rnn_states_actor, masks, action_masks=None, deterministic=False)

evaluate_actions(critic_obs, obs, rnn_states_actor, rnn_states_critic, action, masks, action_masks=None, active_masks=None, critic_masks_batch=None)

get_actions(critic_obs, obs, rnn_states_actor, rnn_states_critic, masks, action_masks=None, deterministic=False)

get_critic_value_normalizer()

get_model_configs(cfg) → Dict[str, Any]

get_values(critic_obs, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

openrl.modules.rl_module module

class openrl.modules.rl_module.RLModule(cfg, act_space: gym.spaces.box.Box, rank: int = 0, world_size: int = 1, device: Union[str, torch.device] = 'cpu', model_configs: Optional[Dict[str, openrl.modules.model_config.ModelTrainConfig]] = None)

Bases: openrl.modules.base_module.BaseModule

abstract act()

abstract evaluate_actions()

abstract get_actions()

abstract get_critic_value_normalizer()

get_model_configs(cfg) → Dict[str, Any]

abstract get_values()

load_policy(model_path: str) → None

restore(model_dir: str) → None

save(save_dir: str) → None

openrl.modules.sac_module module

class openrl.modules.sac_module.SACModule(cfg, input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, deterministic=True)

evaluate_actions()

evaluate_actor_loss(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, action_masks=None, masks_batch=None)

get_actions(obs, deterministic=True)

get_critic_value_normalizer()

get_q_values(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, action_masks=None, masks_batch=None)

get_values(obs, action, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

openrl.modules.vdn_module module

class openrl.modules.vdn_module.VDNModule(cfg, input_space: gym.spaces.box.Box, act_space: gym.spaces.box.Box, device: Union[str, torch.device] = 'cpu', rank: Optional[int] = None, world_size: Optional[int] = None, model_dict: Optional[Dict[str, Any]] = None)

Bases: openrl.modules.rl_module.RLModule

act(obs, rnn_states_actor, masks, action_masks=None)

evaluate_actions(obs_batch, next_obs_batch, rnn_states_batch, rewards_batch, actions_batch, masks, action_masks=None, masks_batch=None, critic_masks_batch=None)

get_actions(obs, rnn_states, masks, action_masks=None)

get_critic_value_normalizer()

get_values(obs, rnn_states_critic, masks)

static init_rnn_states(rollout_num: int, agent_num: int, rnn_layers: int, hidden_size: int)

lr_decay(episode, episodes)

Module contents