Multigrid_learning
Multigrid_learning copied to clipboard
Unclear why tf.Variable is used instead of tf.get_variable in places where there is an option for Variable sharing.
In the function build_Q of model.py, there is a reuse parameter to enable Variable sharing by using tf.variable_scope
. The statement Q = tf.Variable(...)
, however, would create a new Variable every time build_Q
is called regardless of the defined scope. source