Update mnist_dist.py using tensorflow1.6 #247
Conversation
tf.train.Supervisor is deprecated in tensorflow1.6, so I update it using tf.train.MonitoredTrainingSession.
|
Thank you for submitting this pull request, however I do not see a valid CLA on file for you. Before we can merge this request please visit https://yahoocla.herokuapp.com/ and agree to the terms. Thanks! 😄 |
wuyifan18
changed the title
|
Thanks for your contribution! Can you make sure to sign the CLA license? |
|
Thanks for your support! I have signed the CLA license, it works now. |
examples/mnist/spark/mnist_dist.py
Outdated
| is_chief=(task_index == 0), | ||
| checkpoint_dir=logdir, | ||
| hooks=hooks, | ||
| ) as mon_sess: |
There was a problem hiding this comment.
This code throws an error for me (when trying the Standalone instructions). Also, this wouldn't work for args.mode == "inference". Can you update accordingly?
There was a problem hiding this comment.
Can you show me the error？
There was a problem hiding this comment.
File "/Users/leewyang/devpub/TensorFlowOnSpark/examples/mnist/spark/mnist_spark.py", line 18, in <module>
import mnist_dist
File "/Users/leewyang/devpub/TensorFlowOnSpark/examples/mnist/spark/mnist_dist.py", line 117
else:
^
IndentationError: expected an indented block
There was a problem hiding this comment.
I'am coming! I have fix some problems and I can run successfully on Hadoop Cluster.
Meanwhile, I also update the usage of Tensorboard. ：)
3 participants
tf.train.Supervisor is deprecated in tensorflow1.6, so I update it using tf.train.MonitoredTrainingSession. Meanwhile, I update the usage of Tensorboard.