feature(zc): add plan diffuser

ding/entry/serial_entry_offline.py

@@ -75,6 +75,8 @@ def serial_pipeline_offline(
     evaluator_env.seed(cfg.seed, dynamic_seed=False)
     set_pkg_seed(cfg.seed, use_cuda=cfg.policy.cuda)
     policy = create_policy(cfg.policy, model=model, enable_field=['learn', 'eval'])
+    if cfg.policy.collect.data_type == 'diffuser_traj':
+        policy.init_data_normalizer(dataset.normalizer)
 
     if hasattr(policy, 'set_statistic'):
         # useful for setting action bounds for ibc

ding/policy/command_mode_policy_instance.py

@@ -50,6 +50,7 @@
 from .bdq import BDQPolicy
 from .bcq import BCQPolicy
 from .edac import EDACPolicy
+from .plan_diffuser import PDPolicy
 
 
 class EpsCommandModePolicy(CommandModePolicy):
@@ -398,6 +399,10 @@ class BCQCommandModelPolicy(BCQPolicy, DummyCommandModePolicy):
 class EDACCommandModelPolicy(EDACPolicy, DummyCommandModePolicy):
     pass
 
+@POLICY_REGISTRY.register('pd_command')
+class PDCommandModelPolicy(PDPolicy, DummyCommandModePolicy):
+    pass
+
 
 @POLICY_REGISTRY.register('bc_command')
 class BCCommandModePolicy(BehaviourCloningPolicy, DummyCommandModePolicy):