diff --git a/rllib/policy/eager_tf_policy.py b/rllib/policy/eager_tf_policy.py
index 40a2c7986..758cfc948 100644
--- a/rllib/policy/eager_tf_policy.py
+++ b/rllib/policy/eager_tf_policy.py
@@ -393,7 +393,7 @@ def build_eager_tf_policy(name,
                 if action_sampler_fn:
                     dist_inputs = None
                     state_out = []
-                    actions, logp = self.action_sampler_fn(
+                    actions, logp = action_sampler_fn(
                         self,
                         self.model,
                         input_dict[SampleBatch.CUR_OBS],