[rllib] support running older version tensorflow(version < 1.5.0) (#3571)

python/ray/rllib/agents/impala/vtrace_policy_graph.py

@@ -168,7 +168,7 @@ class VTracePolicyGraph(LearningRateSchedule, TFPolicyGraph):
             mask = tf.sequence_mask(self.model.seq_lens, max_seq_len)
             mask = tf.reshape(mask, [-1])
         else:
-            mask = tf.ones_like(rewards)
+            mask = tf.ones_like(rewards, dtype=tf.bool)
 
         # Inputs are reshaped from [B * T] => [T - 1, B] for V-trace calc.
         self.loss = VTraceLoss(

python/ray/rllib/agents/ppo/ppo_policy_graph.py

@@ -206,7 +206,7 @@ class PPOPolicyGraph(LearningRateSchedule, TFPolicyGraph):
             mask = tf.sequence_mask(self.model.seq_lens, max_seq_len)
             mask = tf.reshape(mask, [-1])
         else:
-            mask = tf.ones_like(adv_ph)
+            mask = tf.ones_like(adv_ph, dtype=tf.bool)
 
         self.loss_obj = PPOLoss(
             action_space,

python/ray/tune/logger.py

@@ -8,6 +8,7 @@ import logging
 import numpy as np
 import os
 import yaml
+import distutils.version
 
 import ray.cloudpickle as cloudpickle
 from ray.tune.log_sync import get_syncer
@@ -18,8 +19,11 @@ logger = logging.getLogger(__name__)
 
 try:
     import tensorflow as tf
+    use_tf150_api = (distutils.version.LooseVersion(tf.VERSION) >=
+                     distutils.version.LooseVersion("1.5.0"))
 except ImportError:
     tf = None
+    use_tf150_api = True
     logger.warning("Couldn't import TensorFlow - "
                    "disabling TensorBoard logging.")
 
@@ -155,7 +159,11 @@ def to_tf_values(result, path):
     values = []
     for attr, value in result.items():
         if value is not None:
-            if type(value) in [int, float, np.float32, np.float64, np.int32]:
+            if use_tf150_api:
+                type_list = [int, float, np.float32, np.float64, np.int32]
+            else:
+                type_list = [int, float]
+            if type(value) in type_list:
                 values.append(
                     tf.Summary.Value(
                         tag="/".join(path + [attr]), simple_value=value))