diff --git a/ml-agents/mlagents/trainers/components/bc/model.py b/ml-agents/mlagents/trainers/components/bc/model.py
index c68fd5742f..7f57a1ec7e 100644
--- a/ml-agents/mlagents/trainers/components/bc/model.py
+++ b/ml-agents/mlagents/trainers/components/bc/model.py
@@ -74,7 +74,7 @@ def create_loss(self, learning_rate: float, anneal_steps: int) -> None:
                 power=1.0,
             )
         else:
-            self.annealed_learning_rate = learning_rate
+            self.annealed_learning_rate = tf.Variable(learning_rate)
 
         optimizer = tf.train.AdamOptimizer(learning_rate=self.annealed_learning_rate)
         self.update_batch = optimizer.minimize(self.loss)
diff --git a/ml-agents/mlagents/trainers/tests/test_bcmodule.py b/ml-agents/mlagents/trainers/tests/test_bcmodule.py
index 24f7b7efb7..3a26fd9f56 100644
--- a/ml-agents/mlagents/trainers/tests/test_bcmodule.py
+++ b/ml-agents/mlagents/trainers/tests/test_bcmodule.py
@@ -138,6 +138,26 @@ def test_bcmodule_update(mock_env, trainer_config):
     env.close()
 
 
+# Test with constant pretraining learning rate
+@pytest.mark.parametrize(
+    "trainer_config", [ppo_dummy_config(), sac_dummy_config()], ids=["ppo", "sac"]
+)
+@mock.patch("mlagents.envs.environment.UnityEnvironment")
+def test_bcmodule_constant_lr_update(mock_env, trainer_config):
+    mock_brain = mb.create_mock_3dball_brain()
+    trainer_config["pretraining"]["steps"] = 0
+    env, policy = create_policy_with_bc_mock(
+        mock_env, mock_brain, trainer_config, False, "test.demo"
+    )
+    stats = policy.bc_module.update()
+    for _, item in stats.items():
+        assert isinstance(item, np.float32)
+    old_learning_rate = policy.bc_module.current_lr
+
+    stats = policy.bc_module.update()
+    assert old_learning_rate == policy.bc_module.current_lr
+
+
 # Test with RNN
 @pytest.mark.parametrize(
     "trainer_config", [ppo_dummy_config(), sac_dummy_config()], ids=["ppo", "sac"]