Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

module 'tensorflow' has no attribute 'variable_scope'

See original GitHub issue

Hi I updated tensorflow but get this error for no-limit Texas Holdem:

AttributeError: module 'tensorflow' has no attribute 'variable_scope'

Seems like a migration issue? https://github.com/tensorflow/tensorflow/issues/7285 Thanks for any advice or fixes.

Best, Sy

Issue Analytics

State:
Created 3 years ago
Comments:11 (4 by maintainers)

Top GitHub Comments

11reactions

dipampatel18commented, Sep 6, 2020

Replace tf.variable_scope with tf.compat.v1.variable_scope

3reactions

daochenzhacommented, Mar 31, 2020

@systats Hi, the main reason is wrong inputs to DQN.

global_step = tf$Variable(0, name='global_step', trainable=F) should be global_step = tf$Variable(0L, name='global_step', trainable=F) to make sure that the step is integer
In state_shape=env$state_shape,, RLCard needs a list of integers. However, the R somehow interprets it as a single value. To solve this issue, I transform int to list in DQN so that the single value will also work.
In mlp_layers=c(512, 512), the R interfaces input a float value. However, RLCard requires 512 to be an integer. To solve this issue, I transform mlp_layers to a list of integers in DQN.

If you reclone and reinstall the code of RLCard in the master branch. The script should be able to run.