xbpeng / awr

Implementation of advantage-weighted regression.
MIT License
178 stars 36 forks source link

Upper bound on TensorFlow version #1

Closed tchlux closed 4 years ago

tchlux commented 4 years ago

The current code fails with TensorFlow 2.0, see error message below. For now, this version cap should patch the issue.

Traceback (most recent call last):
  File "run.py", line 9, in <module>
    import learning.awr_agent as awr_agent
  File "awr/learning/awr_agent.py", line 6, in <module>
    import util.net_util as net_util
  File "awr/util/net_util.py", line 5, in <module>
    weight_init=tf.contrib.layers.xavier_initializer(),
AttributeError: module 'tensorflow' has no attribute 'contrib'
xbpeng commented 4 years ago

thanks!