fixing tensorflow tests

ml-agents/mlagents/trainers/policy/tf_policy.py

             mask = np.ones(
                 (
                     len(batched_step_result),
-                    sum(self.behavior_spec.discrete_action_branches),
+                    sum(self.action_spec.discrete_action_branches),
                 ),
                 dtype=np.float32,
             )
             self.mask = tf.cast(self.mask_input, tf.int32)
 
             tf.Variable(
-                int(self.behavior_spec.is_action_continuous()),
+                int(self.action_spec.is_action_continuous()),
                 name="is_continuous_control",
                 trainable=False,
                 dtype=tf.int32,
             tf.Variable(
                 self.m_size, name="memory_size", trainable=False, dtype=tf.int32
             )
-            if self.behavior_spec.is_action_continuous():
+            if self.action_spec.is_action_continuous():
                 tf.Variable(
                     self.act_size[0],
                     name="action_output_shape",

ml-agents/mlagents/trainers/tests/tensorflow/test_models.py

 
 from mlagents.trainers.tf.models import ModelUtils
 from mlagents.tf_utils import tf
-from mlagents_envs.base_env import BehaviorSpec, ActionType
+from mlagents_envs.base_env import BehaviorSpec, ActionSpec
-        ActionType.DISCRETE,
-        (1,),
+        ActionSpec(0, (1,)),
     )
     return behavior_spec
 

ml-agents/mlagents/trainers/tf/components/bc/model.py

         self.done_expert = tf.placeholder(shape=[None, 1], dtype=tf.float32)
         self.done_policy = tf.placeholder(shape=[None, 1], dtype=tf.float32)
 
-        if self.policy.behavior_spec.is_action_continuous():
+        if self.policy.action_spec.is_action_continuous():
             action_length = self.policy.act_size[0]
             self.action_in_expert = tf.placeholder(
                 shape=[None, action_length], dtype=tf.float32

ml-agents/mlagents/trainers/tf/components/reward_signals/curiosity/model.py

         """
         combined_input = tf.concat([encoded_state, encoded_next_state], axis=1)
         hidden = tf.layers.dense(combined_input, 256, activation=ModelUtils.swish)
-        if self.policy.behavior_spec.is_action_continuous():
+        if self.policy.action_spec.is_action_continuous():
             pred_action = tf.layers.dense(
                 hidden, self.policy.act_size[0], activation=None
             )

ml-agents/mlagents/trainers/tf/components/reward_signals/gail/model.py

         self.done_expert = tf.expand_dims(self.done_expert_holder, -1)
         self.done_policy = tf.expand_dims(self.done_policy_holder, -1)
 
-        if self.policy.behavior_spec.is_action_continuous():
+        if self.policy.action_spec.is_action_continuous():
             action_length = self.policy.act_size[0]
             self.action_in_expert = tf.placeholder(
                 shape=[None, action_length], dtype=tf.float32