Remove phrase 'Abstract' from abstract class

algorithms/a2c/agent.py

@@ -16,12 +16,12 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class A2CAgent(Agent):
     """1-Step Advantage Actor-Critic interacting with environment.
 
     Attributes:
@@ -55,7 +55,7 @@ def __init__(
             optims (tuple): optimizers for actor and critic
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.critic = models
         self.actor_optimizer, self.critic_optimizer = optims
@@ -158,7 +158,7 @@ def save_params(self, n_episode: int):
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, score: int, policy_loss: float, value_loss: float):
         total_loss = policy_loss + value_loss

algorithms/bc/ddpg_agent.py

@@ -16,20 +16,20 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import OUNoise
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class BCDDPGAgent(DDPGAgent):
     """BC with DDPG agent interacting with environment.
 
     Attributes:
-        HER (AbstractHER): hinsight experience replay
+        her (HER): hinsight experience replay
         transitions_epi (list): transitions per episode (for HER)
         desired_state (np.ndarray): desired state of current episode
         memory (ReplayBuffer): replay memory
@@ -47,14 +47,14 @@ def __init__(
         models: tuple,
         optims: tuple,
         noise: OUNoise,
-        HER: AbstractHER,
+        her: HER,
     ):
         """Initialization.
         Args:
-            HER (AbstractHER): hinsight experience replay
+            her (HER): hinsight experience replay
 
         """
-        self.HER = HER
+        self.her = her
         DDPGAgent.__init__(self, env, args, hyper_params, models, optims, noise)
 
     # pylint: disable=attribute-defined-outside-init
@@ -66,7 +66,6 @@ def _initialize(self):
 
         # HER
         if self.hyper_params["USE_HER"]:
-            self.her = self.HER()
             if self.hyper_params["DESIRED_STATES_FROM_DEMO"]:
                 self.her.fetch_desired_states_from_demo(demo)
 

algorithms/bc/sac_agent.py

@@ -18,19 +18,19 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.her import HER as AbstractHER
+from algorithms.common.abstract.her import HER
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.sac.agent import Agent as SACAgent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(SACAgent):
+class BCSACAgent(SACAgent):
     """BC with SAC agent interacting with environment.
 
     Attrtibutes:
-        HER (AbstractHER): hinsight experience replay
+        her (HER): hinsight experience replay
         transitions_epi (list): transitions per episode (for HER)
         desired_state (np.ndarray): desired state of current episode
         memory (ReplayBuffer): replay memory
@@ -48,14 +48,14 @@ def __init__(
         models: tuple,
         optims: tuple,
         target_entropy: float,
-        HER: AbstractHER,
+        her: HER,
     ):
         """Initialization.
         Args:
-            HER (AbstractHER): hinsight experience replay
+            her (HER): hinsight experience replay
 
         """
-        self.HER = HER
+        self.her = her
         SACAgent.__init__(self, env, args, hyper_params, models, optims, target_entropy)
 
     # pylint: disable=attribute-defined-outside-init
@@ -67,7 +67,6 @@ def _initialize(self):
 
         # HER
         if self.hyper_params["USE_HER"]:
-            self.her = self.HER()
             if self.hyper_params["DESIRED_STATES_FROM_DEMO"]:
                 self.her.fetch_desired_states_from_demo(demo)
 

algorithms/ddpg/agent.py

@@ -16,15 +16,15 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import OUNoise
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class DDPGAgent(Agent):
     """ActorCritic interacting with environment.
 
     Attributes:
@@ -64,7 +64,7 @@ def __init__(
             noise (OUNoise): random noise for exploration
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.actor_target, self.critic, self.critic_target = models
         self.actor_optimizer, self.critic_optimizer = optims
@@ -196,7 +196,7 @@ def save_params(self, n_episode: int):
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, loss: np.ndarray, score: int):
         """Write log about loss and score"""

algorithms/dqn/agent.py

@@ -22,7 +22,7 @@
 from torch.nn.utils import clip_grad_norm_
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBuffer
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
@@ -31,7 +31,7 @@
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class DQNAgent(Agent):
     """DQN interacting with environment.
 
     Attribute:
@@ -69,7 +69,7 @@ def __init__(
             optim (torch.optim.Adam): optimizers for dqn
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.use_n_step = hyper_params["N_STEP"] > 1
         self.epsilon = hyper_params["MAX_EPSILON"]
@@ -261,7 +261,7 @@ def save_params(self, n_episode: int):
             "dqn_optim_state_dict": self.dqn_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(self, i: int, loss: np.ndarray, score: float):
         """Write log about loss and score"""

algorithms/fd/ddpg_agent.py

@@ -17,12 +17,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class DDPGfDAgent(DDPGAgent):
     """ActorCritic interacting with environment.
 
     Attributes:

algorithms/fd/dqn_agent.py

@@ -22,12 +22,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.dqn.agent import Agent as DQNAgent
+from algorithms.dqn.agent import DQNAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DQNAgent):
+class DQNfDAgent(DQNAgent):
     """DQN interacting with environment.
 
     Attribute:

algorithms/fd/sac_agent.py

@@ -18,12 +18,12 @@
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBufferfD
 from algorithms.common.buffer.replay_buffer import NStepTransitionBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.sac.agent import Agent as SACAgent
+from algorithms.sac.agent import SACAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(SACAgent):
+class SACfDAgent(SACAgent):
     """SAC agent interacting with environment.
 
     Attrtibutes:

algorithms/per/ddpg_agent.py

@@ -13,12 +13,12 @@
 
 from algorithms.common.buffer.priortized_replay_buffer import PrioritizedReplayBuffer
 import algorithms.common.helper_functions as common_utils
-from algorithms.ddpg.agent import Agent as DDPGAgent
+from algorithms.ddpg.agent import DDPGAgent
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(DDPGAgent):
+class DDPGPERAgent(DDPGAgent):
     """ActorCritic interacting with environment.
 
     Attributes:

algorithms/ppo/agent.py

@@ -16,14 +16,14 @@
 import torch.nn as nn
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.env.multiprocessing_env import SubprocVecEnv
 import algorithms.ppo.utils as ppo_utils
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class PPOAgent(Agent):
     """PPO Agent.
 
     Attributes:
@@ -64,7 +64,7 @@ def __init__(
             optims (tuple): optimizers for actor and critic
 
         """
-        AbstractAgent.__init__(self, env_single, args)
+        Agent.__init__(self, env_single, args)
 
         if not self.args.test:
             self.env = env_multi
@@ -251,7 +251,7 @@ def save_params(self, n_episode: int):
             "actor_optim_state_dict": self.actor_optimizer.state_dict(),
             "critic_optim_state_dict": self.critic_optimizer.state_dict(),
         }
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self,

algorithms/sac/agent.py

@@ -18,14 +18,14 @@
 import torch.optim as optim
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class SACAgent(Agent):
     """SAC agent interacting with environment.
 
     Attrtibutes:
@@ -71,7 +71,7 @@ def __init__(
             target_entropy (float): target entropy for the inequality constraint
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.vf, self.vf_target, self.qf_1, self.qf_2 = models
         self.actor_optimizer, self.vf_optimizer = optims[0:2]
@@ -277,7 +277,7 @@ def save_params(self, n_episode: int):
         if self.hyper_params["AUTO_ENTROPY_TUNING"]:
             params["alpha_optim"] = self.alpha_optimizer.state_dict()
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self, i: int, loss: np.ndarray, score: float = 0.0, delayed_update: int = 1

algorithms/td3/agent.py

@@ -16,15 +16,15 @@
 import torch.nn.functional as F
 import wandb
 
-from algorithms.common.abstract.agent import Agent as AbstractAgent
+from algorithms.common.abstract.agent import Agent
 from algorithms.common.buffer.replay_buffer import ReplayBuffer
 import algorithms.common.helper_functions as common_utils
 from algorithms.common.noise import GaussianNoise
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
 
-class Agent(AbstractAgent):
+class TD3Agent(Agent):
     """ActorCritic interacting with environment.
 
     Attributes:
@@ -66,7 +66,7 @@ def __init__(
             noise (GaussianNoise): random noise for exploration
 
         """
-        AbstractAgent.__init__(self, env, args)
+        Agent.__init__(self, env, args)
 
         self.actor, self.actor_target = models[0:2]
         self.critic_1, self.critic_2 = models[2:4]
@@ -225,7 +225,7 @@ def save_params(self, n_episode: int):
             "critic_optim": self.critic_optimizer.state_dict(),
         }
 
-        AbstractAgent.save_params(self, params, n_episode)
+        Agent.save_params(self, params, n_episode)
 
     def write_log(
         self, i: int, loss: np.ndarray, score: float = 0.0, delayed_update: int = 1

examples/lunarlander_continuous_v2/a2c.py

@@ -11,7 +11,7 @@
 import torch
 import torch.optim as optim
 
-from algorithms.a2c.agent import Agent
+from algorithms.a2c.agent import A2CAgent
 from algorithms.common.networks.mlp import MLP, GaussianDist
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -65,7 +65,7 @@ def run(env: gym.Env, args: argparse.Namespace, state_dim: int, action_dim: int)
     optims = (actor_optim, critic_optim)
 
     # create an agent
-    agent = Agent(env, args, hyper_params, models, optims)
+    agent = A2CAgent(env, args, hyper_params, models, optims)
 
     # run
     if args.test: