wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-06 05:16:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source
This commit is contained in:
adoda
2018-12-20 12:27:24 +08:00
committed by Richard Liaw
parent a5309bec7c
commit cf0c4745f4
3 changed files with 11 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/rllib/agents/impala/vtrace_policy_graph.py
+1 -1
View File
@@ -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
+1 -1
View File
@@ -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,