fix(zjow): fix typo for QAC class (#729)

ding/bonus/ddpg.py

@@ -15,7 +15,7 @@
 from ding.utils import set_pkg_seed
 from ding.utils import get_env_fps, render
 from ding.config import save_config_py, compile_config
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.data import DequeBuffer
 from ding.bonus.common import TrainingReturn, EvalReturn
 from ding.config.example.DDPG import supported_env_cfg
@@ -74,7 +74,7 @@ def __init__(
             os.makedirs(self.exp_name)
         save_config_py(self.cfg, os.path.join(self.exp_name, 'policy_config.py'))
         if model is None:
-            model = QAC(**self.cfg.policy.model)
+            model = ContinuousQAC(**self.cfg.policy.model)
         self.buffer_ = DequeBuffer(size=self.cfg.policy.other.replay_buffer.replay_buffer_size)
         self.policy = DDPGPolicy(self.cfg.policy, model=model)
         if policy_state_dict is not None:

ding/bonus/sac.py

@@ -15,7 +15,7 @@
 from ding.utils import set_pkg_seed
 from ding.utils import get_env_fps, render
 from ding.config import save_config_py, compile_config
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.model import model_wrap
 from ding.data import DequeBuffer
 from ding.bonus.common import TrainingReturn, EvalReturn
@@ -75,7 +75,7 @@ def __init__(
             os.makedirs(self.exp_name)
         save_config_py(self.cfg, os.path.join(self.exp_name, 'policy_config.py'))
         if model is None:
-            model = QAC(**self.cfg.policy.model)
+            model = ContinuousQAC(**self.cfg.policy.model)
         self.buffer_ = DequeBuffer(size=self.cfg.policy.other.replay_buffer.replay_buffer_size)
         self.policy = SACPolicy(self.cfg.policy, model=model)
         if policy_state_dict is not None:

ding/bonus/td3.py

@@ -15,7 +15,7 @@
 from ding.utils import set_pkg_seed
 from ding.utils import get_env_fps, render
 from ding.config import save_config_py, compile_config
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.data import DequeBuffer
 from ding.bonus.common import TrainingReturn, EvalReturn
 from ding.config.example.TD3 import supported_env_cfg
@@ -74,7 +74,7 @@ def __init__(
             os.makedirs(self.exp_name)
         save_config_py(self.cfg, os.path.join(self.exp_name, 'policy_config.py'))
         if model is None:
-            model = QAC(**self.cfg.policy.model)
+            model = ContinuousQAC(**self.cfg.policy.model)
         self.buffer_ = DequeBuffer(size=self.cfg.policy.other.replay_buffer.replay_buffer_size)
         self.policy = TD3Policy(self.cfg.policy, model=model)
         if policy_state_dict is not None:

ding/example/collect_demo_data.py

@@ -1,7 +1,7 @@
 import gym
 from ditk import logging
 import torch
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import SACPolicy
 from ding.envs import DingEnvWrapper, BaseEnvManagerV2
 from ding.data import offline_data_save_type
@@ -22,7 +22,7 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
         policy = SACPolicy(cfg.policy, model=model, enable_field=['collect'])
         state_dict = torch.load(cfg.policy.collect.state_dict_path, map_location='cpu')
         policy.collect_mode.load_state_dict(state_dict)

ding/example/ddpg.py

@@ -1,6 +1,6 @@
 import gym
 from ditk import logging
-from ding.model.template.qac import QAC
+from ding.model.template.qac import ContinuousQAC
 from ding.policy import DDPGPolicy
 from ding.envs import DingEnvWrapper, BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -27,7 +27,7 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
         buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
         policy = DDPGPolicy(cfg.policy, model=model)
 

ding/example/sac.py

@@ -1,5 +1,5 @@
 from ditk import logging
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import SACPolicy
 from ding.envs import BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -27,7 +27,7 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
         buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
         policy = SACPolicy(cfg.policy, model=model)
 

ding/example/sqil_continuous.py

@@ -1,6 +1,6 @@
 from ditk import logging
 import torch
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import SQILSACPolicy
 from ding.envs import BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -35,8 +35,8 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
-        expert_model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
+        expert_model = ContinuousQAC(**cfg.policy.model)
 
         buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
         expert_buffer = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)

ding/example/td3.py

@@ -1,5 +1,5 @@
 from ditk import logging
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import TD3Policy
 from ding.envs import BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -27,7 +27,7 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
         buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
         policy = TD3Policy(cfg.policy, model=model)
 

ding/model/template/maqac.py

@@ -145,12 +145,12 @@ def compute_actor(self, inputs: Dict) -> Dict:
             - q_value (:obj:`torch.FloatTensor`): :math:`(B, )`, B is batch size.
         Examples:
             >>> # Regression mode
-            >>> model = QAC(64, 64, 'regression')
+            >>> model = DiscreteQAC(64, 64, 'regression')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> assert actor_outputs['action'].shape == torch.Size([4, 64])
             >>> # Reparameterization Mode
-            >>> model = QAC(64, 64, 'reparameterization')
+            >>> model = DiscreteQAC(64, 64, 'reparameterization')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> actor_outputs['logit'][0].shape # mu
@@ -329,12 +329,12 @@ def forward(self, inputs: Union[torch.Tensor, Dict], mode: str) -> Dict:
 
         Actor Examples:
             >>> # Regression mode
-            >>> model = QAC(64, 64, 'regression')
+            >>> model = ContinuousQAC(64, 64, 'regression')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> assert actor_outputs['action'].shape == torch.Size([4, 64])
             >>> # Reparameterization Mode
-            >>> model = QAC(64, 64, 'reparameterization')
+            >>> model = ContinuousQAC(64, 64, 'reparameterization')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> actor_outputs['logit'][0].shape # mu
@@ -376,12 +376,12 @@ def compute_actor(self, inputs: Dict) -> Dict:
                 (action_args are continuous real value)
         Examples:
             >>> # Regression mode
-            >>> model = QAC(64, 64, 'regression')
+            >>> model = ContinuousQAC(64, 64, 'regression')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> assert actor_outputs['action'].shape == torch.Size([4, 64])
             >>> # Reparameterization Mode
-            >>> model = QAC(64, 64, 'reparameterization')
+            >>> model = ContinuousQAC(64, 64, 'reparameterization')
             >>> inputs = torch.randn(4, 64)
             >>> actor_outputs = model(inputs,'compute_actor')
             >>> actor_outputs['logit'][0].shape # mu

dizoo/classic_control/pendulum/entry/pendulum_ddpg_main.py

@@ -6,7 +6,7 @@
 from ding.worker import BaseLearner, SampleSerialCollector, InteractionSerialEvaluator, AdvancedReplayBuffer
 from ding.envs import BaseEnvManager, DingEnvWrapper
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from dizoo.classic_control.pendulum.envs import PendulumEnv
 from dizoo.classic_control.pendulum.config.pendulum_ddpg_config import pendulum_ddpg_config
@@ -50,7 +50,7 @@ def main(cfg, seed=0):
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
     # Set up RL Policy
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
 
     # Set up collection, training and evaluation utilities

dizoo/classic_control/pendulum/entry/pendulum_td3_main.py

@@ -7,7 +7,7 @@
 from ding.worker import BaseLearner, SampleSerialCollector, InteractionSerialEvaluator, AdvancedReplayBuffer
 from ding.envs import BaseEnvManager, DingEnvWrapper
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from dizoo.classic_control.pendulum.envs import PendulumEnv
 from dizoo.classic_control.pendulum.config.pendulum_td3_config import pendulum_td3_config
@@ -40,7 +40,7 @@ def main(cfg, seed=0):
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
     # Set up RL Policy
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
     # lr_scheduler demo
     lr_scheduler = LambdaLR(

dizoo/dmc2gym/entry/dmc2gym_sac_pixel_main.py

@@ -2,7 +2,7 @@
 from ditk import logging
 import os
 import numpy as np
-from ding.model.template.qac import QAC
+from ding.model.template.qac import ContinuousQAC
 from ding.policy import SACPolicy
 from ding.envs import BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -36,7 +36,7 @@ def main():
 
             set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-            model = QAC(**cfg.policy.model)
+            model = ContinuousQAC(**cfg.policy.model)
             logging.info(model)
             buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
             policy = SACPolicy(cfg.policy, model=model)

dizoo/dmc2gym/entry/dmc2gym_sac_state_main.py

@@ -1,5 +1,5 @@
 from ditk import logging
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import SACPolicy
 from ding.envs import BaseEnvManagerV2
 from ding.data import DequeBuffer
@@ -36,7 +36,7 @@ def main():
 
             set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-            model = QAC(**cfg.policy.model)
+            model = ContinuousQAC(**cfg.policy.model)
             buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
             policy = SACPolicy(cfg.policy, model=model)
 

dizoo/gym_hybrid/entry/gym_hybrid_ddpg_eval.py

@@ -10,7 +10,7 @@
 from ding.envs import BaseEnvManager, DingEnvWrapper
 from ding.envs import get_vec_env_setting
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from ding.rl_utils import get_epsilon_greedy_fn
 from dizoo.gym_hybrid.config.gym_hybrid_ddpg_config import gym_hybrid_ddpg_config, gym_hybrid_ddpg_create_config
@@ -33,7 +33,7 @@ def main(main_cfg, create_cfg, seed=0):
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
     # Set up RL Policy
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
     policy.eval_mode.load_state_dict(torch.load(cfg.policy.load_path, map_location='cpu'))
 

dizoo/gym_hybrid/entry/gym_hybrid_ddpg_main.py

@@ -7,7 +7,7 @@
 from ding.worker import BaseLearner, SampleSerialCollector, InteractionSerialEvaluator, AdvancedReplayBuffer
 from ding.envs import BaseEnvManager
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from ding.rl_utils import get_epsilon_greedy_fn
 from dizoo.gym_hybrid.envs.gym_hybrid_env import GymHybridEnv
@@ -43,7 +43,7 @@ def main(cfg, seed=0):
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
     # Set up RL Policy
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
 
     # Set up collection, training and evaluation utilities

dizoo/metadrive/config/metadrive_onppo_config.py

@@ -5,7 +5,7 @@
 import gym
 from ding.envs import BaseEnvManager, SyncSubprocessEnvManager
 from ding.config import compile_config
-from ding.model.template import QAC, VAC
+from ding.model.template import ContinuousQAC, VAC
 from ding.policy import PPOPolicy
 from ding.worker import SampleSerialCollector, InteractionSerialEvaluator, BaseLearner
 from dizoo.metadrive.env.drive_env import MetaDrivePPOOriginEnv

dizoo/mujoco/entry/mujoco_ddpg_eval.py

@@ -10,7 +10,7 @@
 from ding.envs import BaseEnvManager
 from ding.envs import get_vec_env_setting, create_env_manager
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from ding.rl_utils import get_epsilon_greedy_fn
 from dizoo.mujoco.envs.mujoco_env import MujocoEnv
@@ -41,7 +41,7 @@ def main(main_cfg, seed=0):
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
     # Set up RL Policy
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
     policy.eval_mode.load_state_dict(torch.load(cfg.policy.load_path, map_location='cpu'))
 

dizoo/mujoco/entry/mujoco_ddpg_main.py

@@ -7,7 +7,7 @@
 from ding.worker import BaseLearner, SampleSerialCollector, InteractionSerialEvaluator, AdvancedReplayBuffer
 from ding.envs import BaseEnvManager, DingEnvWrapper
 from ding.policy import DDPGPolicy
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.utils import set_pkg_seed
 from dizoo.mujoco.envs.mujoco_env import MujocoEnv
 from dizoo.mujoco.config.hopper_ddpg_config import hopper_ddpg_config
@@ -37,7 +37,7 @@ def main(cfg, seed=0, max_iterations=int(1e10)):
     evaluator_env.seed(seed, dynamic_seed=False)
     set_pkg_seed(seed, use_cuda=cfg.policy.cuda)
 
-    model = QAC(**cfg.policy.model)
+    model = ContinuousQAC(**cfg.policy.model)
     policy = DDPGPolicy(cfg.policy, model=model)
     tb_logger = SummaryWriter(os.path.join('./log/', 'serial'))
     learner = BaseLearner(cfg.policy.learn.learner, policy.learn_mode, tb_logger)

dizoo/mujoco/example/mujoco_sac.py

@@ -1,5 +1,5 @@
 from ditk import logging
-from ding.model import QAC
+from ding.model import ContinuousQAC
 from ding.policy import SACPolicy
 from ding.envs import DingEnvWrapper, SubprocessEnvManagerV2
 from ding.data import DequeBuffer
@@ -26,7 +26,7 @@ def main():
 
         set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
 
-        model = QAC(**cfg.policy.model)
+        model = ContinuousQAC(**cfg.policy.model)
         buffer_ = DequeBuffer(size=cfg.policy.other.replay_buffer.replay_buffer_size)
         policy = SACPolicy(cfg.policy, model=model)