polish(zc): change PD config name (#749)

* add action

* change entry

Author: Super1ce

ding/policy/plan_diffuser.py

@@ -178,7 +178,7 @@ def _init_learn(self) -> None:
         self.step_start_update_target = self._cfg.learn.step_start_update_target
         self.target_weight = self._cfg.learn.target_weight
         self.value_step = self._cfg.learn.value_step
-        self.use_target = True
+        self.use_target = False
         self.horizon = self._cfg.model.diffuser_model_cfg.horizon
         self.include_returns = self._cfg.learn.include_returns
 

ding/utils/data/dataset.py

@@ -1090,11 +1090,13 @@ def __getitem__(self, idx, eps=1e-4):
                 'trajectories': trajectories,
                 'returns': returns,
                 'done': done,
+                'action': actions,
             }
         else:
             batch = {
                 'trajectories': trajectories,
                 'done': done,
+                'action': actions,
             }
 
         batch.update(self.get_conditions(observations))

dizoo/d4rl/config/antmaze_umaze_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=37,
                     dim=32,

dizoo/d4rl/config/halfcheetah_medium_expert_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=23,
                     dim=32,

dizoo/d4rl/config/halfcheetah_medium_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=23,
                     dim=32,

dizoo/d4rl/config/hopper_medium_expert_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=14,
                     dim=32,

dizoo/d4rl/config/hopper_medium_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=14,
                     dim=32,

dizoo/d4rl/config/walker2d_medium_expert_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=23,
                     dim=32,

dizoo/d4rl/config/walker2d_medium_pd_config.py

@@ -24,7 +24,7 @@
         model=dict(
             diffuser_model='GaussianDiffusion',
             diffuser_model_cfg=dict(
-                model='TemporalUnet',
+                model='DiffusionUNet1d',
                 model_cfg=dict(
                     transition_dim=23,
                     dim=32,

dizoo/d4rl/entry/d4rl_pd_main.py

@@ -16,6 +16,6 @@ def train(args):
 
     parser = argparse.ArgumentParser()
     parser.add_argument('--seed', '-s', type=int, default=10)
-    parser.add_argument('--config', '-c', type=str, default='hopper_expert_cql_config.py')
+    parser.add_argument('--config', '-c', type=str, default='halfcheetah_medium_pd_config.py')
     args = parser.parse_args()
     train(args)