Fix missing learning rate and entropy coeff schedule for torch PPO (r…

…ay-project#8572)
mrkulk · May 23, 2020 · d6f78f5 · d6f78f5
1 parent 2ab1b77
commit d6f78f5
Showing 1 changed file with 4 additions and 1 deletion.
diff --git a/rllib/agents/ppo/ppo_torch_policy.py b/rllib/agents/ppo/ppo_torch_policy.py
@@ -223,4 +223,7 @@ def setup_mixins(policy, obs_space, action_space, config):
     extra_grad_process_fn=apply_grad_clipping,
     before_init=setup_config,
     after_init=setup_mixins,
-    mixins=[KLCoeffMixin, ValueNetworkMixin])
+    mixins=[
+        LearningRateSchedule, EntropyCoeffSchedule, KLCoeffMixin,
+        ValueNetworkMixin
+    ])