Handle gin.bind_parameter in tests (#395)

If a test needs to `gin.bind_parameter` it should also clear that, or
make the binding within a `gin.config_scope`, otherwise the bindings may
leak to other tests if all the tests are run by e.g. pytest.

This patch uses both approaches: for tests that set up bindings in `setUp`
we clear in `tearDown`, else we use the scoped approach.

Issue #394

compiler_opt/distributed/worker_test.py

@@ -30,12 +30,13 @@ def __init__(self, argument):
 class WorkerTest(absltest.TestCase):
 
   def test_gin_args(self):
-    with gin.unlock_config():
-      gin.bind_parameter('_test.SomeType.argument', 42)
-    real_args = worker.get_full_worker_args(
-        SomeType, more_args=2, even_more_args='hi')
-    self.assertDictEqual(real_args,
-                         dict(argument=42, more_args=2, even_more_args='hi'))
+    with gin.config_scope('worker_test'):
+      with gin.unlock_config():
+        gin.bind_parameter('_test.SomeType.argument', 42)
+      real_args = worker.get_full_worker_args(
+          SomeType, more_args=2, even_more_args='hi')
+      self.assertDictEqual(real_args,
+                           dict(argument=42, more_args=2, even_more_args='hi'))
 
 
 if __name__ == '__main__':

compiler_opt/es/blackbox_learner_test.py

@@ -40,6 +40,9 @@
 class BlackboxLearnerTests(absltest.TestCase):
   """Tests for blackbox_learner"""
 
+  def tearDown(self):
+    gin.clear_config()
+
   def setUp(self):
     super().setUp()
 

compiler_opt/rl/agent_config_test.py

@@ -51,34 +51,42 @@ def setUp(self):
     super().setUp()
 
   def test_create_behavioral_cloning_agent(self):
-    gin.bind_parameter('create_agent.policy_network', q_network.QNetwork)
-    gin.bind_parameter('BehavioralCloningAgent.optimizer',
-                       tf.compat.v1.train.AdamOptimizer())
-    tf_agent = agent_config.create_agent(
-        agent_config.BCAgentConfig(
-            time_step_spec=self._time_step_spec, action_spec=self._action_spec),
-        preprocessing_layer_creator=_observation_processing_layer)
-    self.assertIsInstance(tf_agent,
-                          behavioral_cloning_agent.BehavioralCloningAgent)
+    with gin.config_scope('test_create_behavioral_cloning_agent'):
+      gin.bind_parameter('create_agent.policy_network', q_network.QNetwork)
+      gin.bind_parameter('BehavioralCloningAgent.optimizer',
+                         tf.compat.v1.train.AdamOptimizer())
+      tf_agent = agent_config.create_agent(
+          agent_config.BCAgentConfig(
+              time_step_spec=self._time_step_spec,
+              action_spec=self._action_spec),
+          preprocessing_layer_creator=_observation_processing_layer)
+      self.assertIsInstance(tf_agent,
+                            behavioral_cloning_agent.BehavioralCloningAgent)
 
   def test_create_dqn_agent(self):
-    gin.bind_parameter('create_agent.policy_network', q_network.QNetwork)
-    gin.bind_parameter('DqnAgent.optimizer', tf.compat.v1.train.AdamOptimizer())
-    tf_agent = agent_config.create_agent(
-        agent_config.DQNAgentConfig(
-            time_step_spec=self._time_step_spec, action_spec=self._action_spec),
-        preprocessing_layer_creator=_observation_processing_layer)
-    self.assertIsInstance(tf_agent, dqn_agent.DqnAgent)
+    with gin.config_scope('test_create_dqn_agent'):
+      gin.bind_parameter('create_agent.policy_network', q_network.QNetwork)
+      gin.bind_parameter('DqnAgent.optimizer',
+                         tf.compat.v1.train.AdamOptimizer())
+      tf_agent = agent_config.create_agent(
+          agent_config.DQNAgentConfig(
+              time_step_spec=self._time_step_spec,
+              action_spec=self._action_spec),
+          preprocessing_layer_creator=_observation_processing_layer)
+      self.assertIsInstance(tf_agent, dqn_agent.DqnAgent)
 
   def test_create_ppo_agent(self):
-    gin.bind_parameter('create_agent.policy_network',
-                       actor_distribution_network.ActorDistributionNetwork)
-    gin.bind_parameter('PPOAgent.optimizer', tf.compat.v1.train.AdamOptimizer())
-    tf_agent = agent_config.create_agent(
-        agent_config.PPOAgentConfig(
-            time_step_spec=self._time_step_spec, action_spec=self._action_spec),
-        preprocessing_layer_creator=_observation_processing_layer)
-    self.assertIsInstance(tf_agent, ppo_agent.PPOAgent)
+    with gin.config_scope('test_create_ppo_agent'):
+      gin.bind_parameter('create_agent.policy_network',
+                         actor_distribution_network.ActorDistributionNetwork)
+      gin.bind_parameter('PPOAgent.optimizer',
+                         tf.compat.v1.train.AdamOptimizer())
+      tf_agent = agent_config.create_agent(
+          agent_config.PPOAgentConfig(
+              time_step_spec=self._time_step_spec,
+              action_spec=self._action_spec),
+          preprocessing_layer_creator=_observation_processing_layer)
+      self.assertIsInstance(tf_agent, ppo_agent.PPOAgent)
 
 
 if __name__ == '__main__':