[RLlib] tf-eager support for ES and ARS (tf2.x preparation). (#9207)

python/ray/experimental/tf_utils.py

@@ -44,6 +44,7 @@ class TensorFlowVariables:
                 operation to extract all variables from.
             sess (Optional[tf.Session]): Optional tf.Session used for running
                 the get and set methods in tf graph mode.
+                Use None for tf eager.
             input_variables (List[tf.Variables]): Variables to include in the
                 list.
         """
@@ -103,14 +104,6 @@ class TensorFlowVariables:
             for v in variable_list:
                 self.variables[v.name] = v
 
-    def set_session(self, sess):
-        """Sets the current session used by the class.
-
-        Args:
-            sess (tf.Session): Session to set the attribute with.
-        """
-        self.sess = sess
-
     def get_flat_size(self):
         """Returns the total length of all of the flattened variables.
 
@@ -120,22 +113,12 @@ class TensorFlowVariables:
         return sum(
             np.prod(v.get_shape().as_list()) for v in self.variables.values())
 
-    def _check_sess(self):
-        """Checks if the session is set, and if not throw an error message."""
-        if tf1.executing_eagerly():
-            return
-        assert self.sess is not None, \
-            "The session is not set. Set the session either by passing it " \
-            "into the TensorFlowVariables constructor or by calling " \
-            "set_session(sess)."
-
     def get_flat(self):
         """Gets the weights and returns them as a flat array.
 
         Returns:
             1D Array containing the flattened weights.
         """
-        self._check_sess()
         # Eager mode.
         if not self.sess:
             return np.concatenate(
@@ -156,7 +139,6 @@ class TensorFlowVariables:
         Args:
             new_weights (np.ndarray): Flat array containing weights.
         """
-        self._check_sess()
         shapes = [v.get_shape().as_list() for v in self.variables.values()]
         arrays = unflatten(new_weights, shapes)
         if not self.sess:
@@ -176,7 +158,6 @@ class TensorFlowVariables:
         Returns:
             Dictionary mapping variable names to their weights.
         """
-        self._check_sess()
         # Eager mode.
         if not self.sess:
             return self.variables
@@ -194,20 +175,23 @@ class TensorFlowVariables:
             new_weights (Dict): Dictionary mapping variable names to their
                 weights.
         """
-        self._check_sess()
-        assign_list = [
-            self.assignment_nodes[name] for name in new_weights.keys()
-            if name in self.assignment_nodes
-        ]
-        assert assign_list, ("No variables in the input matched those in the "
-                             "network. Possible cause: Two networks were "
-                             "defined in the same TensorFlow graph. To fix "
-                             "this, place each network definition in its own "
-                             "tf.Graph.")
-        self.sess.run(
-            assign_list,
-            feed_dict={
-                self.placeholders[name]: value
-                for (name, value) in new_weights.items()
-                if name in self.placeholders
-            })
+        if self.sess is None:
+            for name, var in self.variables.items():
+                var.assign(new_weights[name])
+        else:
+            assign_list = [
+                self.assignment_nodes[name] for name in new_weights.keys()
+                if name in self.assignment_nodes
+            ]
+            assert assign_list, \
+                "No variables in the input matched those in the network. " \
+                "Possible cause: Two networks were defined in the same " \
+                "TensorFlow graph. To fix this, place each network " \
+                "definition in its own tf.Graph."
+            self.sess.run(
+                assign_list,
+                feed_dict={
+                    self.placeholders[name]: value
+                    for (name, value) in new_weights.items()
+                    if name in self.placeholders
+                })

python/ray/tests/test_tensorflow.py

@@ -122,10 +122,9 @@ def test_tensorflow_variables(ray_start_2_cpus):
     variables2.set_flat(flat_weights)
     assert_almost_equal(flat_weights, variables2.get_flat())
 
-    variables3 = ray.experimental.tf_utils.TensorFlowVariables([loss2])
-    assert variables3.sess is None
     sess = tf.Session()
-    variables3.set_session(sess)
+    variables3 = ray.experimental.tf_utils.TensorFlowVariables(
+        [loss2], sess=sess)
     assert variables3.sess == sess
 
 

rllib/agents/ars/ars_tf_policy.py

@@ -9,11 +9,13 @@ import ray.experimental.tf_utils
 from ray.rllib.agents.es.es_tf_policy import make_session
 from ray.rllib.models import ModelCatalog
 from ray.rllib.policy.sample_batch import SampleBatch
+from ray.rllib.utils import try_import_tree
 from ray.rllib.utils.filter import get_filter
 from ray.rllib.utils.framework import try_import_tf
 from ray.rllib.utils.spaces.space_utils import unbatch
 
 tf1, tf, tfv = try_import_tf()
+tree = try_import_tree()
 
 
 class ARSTFPolicy:
@@ -27,13 +29,17 @@ class ARSTFPolicy:
                                              self.preprocessor.shape)
 
         self.single_threaded = config.get("single_threaded", False)
-        self.sess = make_session(single_threaded=self.single_threaded)
-
-        self.inputs = tf1.placeholder(tf.float32,
-                                      [None] + list(self.preprocessor.shape))
+        if config["framework"] == "tf":
+            self.sess = make_session(single_threaded=self.single_threaded)
+            self.inputs = tf1.placeholder(
+                tf.float32, [None] + list(self.preprocessor.shape))
+        else:
+            if not tf1.executing_eagerly():
+                tf1.enable_eager_execution()
+            self.sess = self.inputs = None
 
         # Policy network.
-        dist_class, dist_dim = ModelCatalog.get_action_dist(
+        self.dist_class, dist_dim = ModelCatalog.get_action_dist(
             self.action_space, config["model"], dist_type="deterministic")
 
         self.model = ModelCatalog.get_model_v2(
@@ -41,18 +47,22 @@ class ARSTFPolicy:
             action_space=self.action_space,
             num_outputs=dist_dim,
             model_config=config["model"])
-        dist_inputs, _ = self.model({SampleBatch.CUR_OBS: self.inputs})
-        dist = dist_class(dist_inputs, self.model)
 
-        self.sampler = dist.sample()
-
-        self.variables = ray.experimental.tf_utils.TensorFlowVariables(
-            dist_inputs, self.sess)
+        self.sampler = None
+        if self.sess:
+            dist_inputs, _ = self.model({SampleBatch.CUR_OBS: self.inputs})
+            dist = self.dist_class(dist_inputs, self.model)
+            self.sampler = dist.sample()
+            self.variables = ray.experimental.tf_utils.TensorFlowVariables(
+                dist_inputs, self.sess)
+            self.sess.run(tf1.global_variables_initializer())
+        else:
+            self.variables = ray.experimental.tf_utils.TensorFlowVariables(
+                [], None, self.model.variables())
 
         self.num_params = sum(
             np.prod(variable.shape.as_list())
             for _, variable in self.variables.variables.items())
-        self.sess.run(tf1.global_variables_initializer())
 
     def compute_actions(self,
                         observation,
@@ -64,12 +74,23 @@ class ARSTFPolicy:
             observation = observation[0]
         observation = self.preprocessor.transform(observation)
         observation = self.observation_filter(observation[None], update=update)
-        action = self.sess.run(
-            self.sampler, feed_dict={self.inputs: observation})
-        action = unbatch(action)
+
+        # `actions` is a list of (component) batches.
+        # Eager mode.
+        if not self.sess:
+            dist_inputs, _ = self.model({SampleBatch.CUR_OBS: observation})
+            dist = self.dist_class(dist_inputs, self.model)
+            actions = dist.sample()
+            actions = tree.map_structure(lambda a: a.numpy(), actions)
+        # Graph mode.
+        else:
+            actions = self.sess.run(
+                self.sampler, feed_dict={self.inputs: observation})
+
+        actions = unbatch(actions)
         if add_noise and isinstance(self.action_space, gym.spaces.Box):
-            action += np.random.randn(*action.shape) * self.action_noise_std
-        return action
+            actions += np.random.randn(*actions.shape) * self.action_noise_std
+        return actions
 
     def compute_single_action(self,
                               observation,

rllib/agents/ars/tests/test_ars.py

@@ -17,7 +17,7 @@ class TestARS(unittest.TestCase):
 
         num_iterations = 2
 
-        for _ in framework_iterator(config, ("tf", "torch")):
+        for _ in framework_iterator(config):
             plain_config = config.copy()
             trainer = ars.ARSTrainer(config=plain_config, env="CartPole-v0")
             for i in range(num_iterations):
@@ -25,6 +25,8 @@ class TestARS(unittest.TestCase):
                 print(results)
 
             check_compute_single_action(trainer)
+            trainer.stop()
+        ray.shutdown()
 
 
 if __name__ == "__main__":

rllib/agents/es/es_tf_policy.py

@@ -76,29 +76,40 @@ class ESTFPolicy:
         self.observation_filter = get_filter(config["observation_filter"],
                                              self.preprocessor.shape)
         self.single_threaded = config.get("single_threaded", False)
-        self.sess = make_session(single_threaded=self.single_threaded)
-        self.inputs = tf1.placeholder(tf.float32,
-                                      [None] + list(self.preprocessor.shape))
+        if config["framework"] == "tf":
+            self.sess = make_session(single_threaded=self.single_threaded)
+            self.inputs = tf1.placeholder(
+                tf.float32, [None] + list(self.preprocessor.shape))
+        else:
+            if not tf1.executing_eagerly():
+                tf1.enable_eager_execution()
+            self.sess = self.inputs = None
 
         # Policy network.
-        dist_class, dist_dim = ModelCatalog.get_action_dist(
+        self.dist_class, dist_dim = ModelCatalog.get_action_dist(
             self.action_space, config["model"], dist_type="deterministic")
+
         self.model = ModelCatalog.get_model_v2(
             obs_space=self.preprocessor.observation_space,
             action_space=action_space,
             num_outputs=dist_dim,
             model_config=config["model"])
-        dist_inputs, _ = self.model({SampleBatch.CUR_OBS: self.inputs})
-        dist = dist_class(dist_inputs, self.model)
-        self.sampler = dist.sample()
 
-        self.variables = ray.experimental.tf_utils.TensorFlowVariables(
-            dist_inputs, self.sess)
+        self.sampler = None
+        if self.sess:
+            dist_inputs, _ = self.model({SampleBatch.CUR_OBS: self.inputs})
+            dist = self.dist_class(dist_inputs, self.model)
+            self.sampler = dist.sample()
+            self.variables = ray.experimental.tf_utils.TensorFlowVariables(
+                dist_inputs, self.sess)
+            self.sess.run(tf1.global_variables_initializer())
+        else:
+            self.variables = ray.experimental.tf_utils.TensorFlowVariables(
+                [], None, self.model.variables())
 
         self.num_params = sum(
             np.prod(variable.shape.as_list())
             for _, variable in self.variables.variables.items())
-        self.sess.run(tf1.global_variables_initializer())
 
     def compute_actions(self,
                         observation,
@@ -111,8 +122,17 @@ class ESTFPolicy:
         observation = self.preprocessor.transform(observation)
         observation = self.observation_filter(observation[None], update=update)
         # `actions` is a list of (component) batches.
-        actions = self.sess.run(
-            self.sampler, feed_dict={self.inputs: observation})
+        # Eager mode.
+        if not self.sess:
+            dist_inputs, _ = self.model({SampleBatch.CUR_OBS: observation})
+            dist = self.dist_class(dist_inputs, self.model)
+            actions = dist.sample()
+            actions = tree.map_structure(lambda a: a.numpy(), actions)
+        # Graph mode.
+        else:
+            actions = self.sess.run(
+                self.sampler, feed_dict={self.inputs: observation})
+
         if add_noise:
             actions = tree.map_structure(self._add_noise, actions,
                                          self.action_space_struct)

rllib/agents/es/tests/test_es.py

@@ -18,7 +18,7 @@ class TestES(unittest.TestCase):
 
         num_iterations = 2
 
-        for _ in framework_iterator(config, ("tf", "torch")):
+        for _ in framework_iterator(config):
             plain_config = config.copy()
             trainer = es.ESTrainer(config=plain_config, env="CartPole-v0")
             for i in range(num_iterations):
@@ -26,6 +26,8 @@ class TestES(unittest.TestCase):
                 print(results)
 
             check_compute_single_action(trainer)
+            trainer.stop()
+        ray.shutdown()
 
 
 if __name__ == "__main__":