From acbe0b4e5f6cff6ce8fe8496d2c225b808dba9b1 Mon Sep 17 00:00:00 2001 From: Jones Wong Date: Fri, 22 Feb 2019 02:47:01 +0800 Subject: [PATCH] Fix twin q bug (#4108) --- python/ray/rllib/agents/ddpg/ddpg_policy_graph.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/python/ray/rllib/agents/ddpg/ddpg_policy_graph.py b/python/ray/rllib/agents/ddpg/ddpg_policy_graph.py index 3032bbb3f..028a2e266 100644 --- a/python/ray/rllib/agents/ddpg/ddpg_policy_graph.py +++ b/python/ray/rllib/agents/ddpg/ddpg_policy_graph.py @@ -147,7 +147,7 @@ class ActorCriticLoss(object): q_t_selected = tf.squeeze(q_t, axis=len(q_t.shape) - 1) if twin_q: - twin_q_t_selected = tf.squeeze(q_t, axis=len(q_t.shape) - 1) + twin_q_t_selected = tf.squeeze(twin_q_t, axis=len(q_t.shape) - 1) q_tp1 = tf.minimum(q_tp1, twin_q_tp1) q_tp1_best = tf.squeeze(input=q_tp1, axis=len(q_tp1.shape) - 1)