gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-17 19:20:19 +01:00
Code Issues Projects Releases Wiki Activity

bug-fix in dueling dqn

Browse Source
This commit is contained in:
Gal Leibovich
2018-08-30 18:02:20 +03:00
parent bbe7ac3338
commit d2623c0eee
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
rl_coach/architectures/tensorflow_components/heads/dueling_q_head.py
View File

@@ -47,7 +47,7 @@ class DuelingQHead(QHead):
with tf.variable_scope("action_advantage"):
action_advantage = self.dense_layer(512)(input_layer, activation=self.activation_function, name='fc1')
action_advantage = self.dense_layer(self.num_actions)(action_advantage, name='fc2')
action_advantage = action_advantage - tf.reduce_mean(action_advantage)
action_advantage = action_advantage - tf.reduce_mean(action_advantage, axis=1)
# merge to state-action value function Q
self.output = tf.add(state_value, action_advantage, name='output')