Right that addresses the error. Thanks!! My original point was that in addition to the uninitialized variable, we could also enhance interactive mode detection - as of now interactive mode is always off for Zeppelin.
_____________________________ From: moon soo Lee <m...@apache.org<mailto:m...@apache.org>> Sent: Friday, March 17, 2017 2:07 PM Subject: Re: Tensorflow not working in Zeppelin 0.7.0 To: <users@zeppelin.apache.org<mailto:users@zeppelin.apache.org>> Saw patch [1] has been made in Tensorflow. Thanks Sriram for taking care of this issue! Best, moon [1] https://github.com/tensorflow/tensorflow/pull/8420 On Wed, Mar 15, 2017 at 8:32 PM Sriram Narayanamoorthy <nsrira...@gmail.com<mailto:nsrira...@gmail.com>> wrote: We fixed the issue on tensorflow by a simply initializing the _interactive variable in the logging file. And I tested it out on both Zeppelin and Jupyter and it seems to work. Do you have a better version of this in mind? On Tue, Mar 14, 2017 at 8:45 PM Felix Cheung <felixcheun...@hotmail.com<mailto:felixcheun...@hotmail.com>> wrote: I realize there is a discussion on be Tensorflow issue, aside from the thought of initializing the variables, I think they should broaden their detection of interactive shell to include Zeppelin, Jupyter, etc. ________________________________ From: Sriram Narayanamoorthy <nsrira...@gmail.com<mailto:nsrira...@gmail.com>> Sent: Tuesday, March 14, 2017 4:34:13 PM To: users@zeppelin.apache.org<mailto:users@zeppelin.apache.org> Subject: Re: Tensorflow not working in Zeppelin 0.7.0 Python 3.5 and Tensorflow 1.0 On Tue, Mar 14, 2017 at 4:21 PM, Jianfeng (Jeff) Zhang<jzh...@hortonworks.com<mailto:jzh...@hortonworks.com>> wrote: Which version of tensorflow and python do you use ? I believe I use it before and works for me. Best Regard, Jeff Zhang From: Sriram Narayanamoorthy <nsrira...@gmail.com<mailto:nsrira...@gmail.com>> Reply-To: "users@zeppelin.apache.org<mailto:users@zeppelin.apache.org>" <users@zeppelin.apache.org<mailto:users@zeppelin.apache.org>> Date: Wednesday, March 15, 2017 at 2:19 AM To: "users@zeppelin.apache.org<mailto:users@zeppelin.apache.org>" <users@zeppelin.apache.org<mailto:users@zeppelin.apache.org>> Subject: Re: Tensorflow not working in Zeppelin 0.7.0 Thanks. I have submitted a Jira ticket: https://issues.apache.org/jira/browse/ZEPPELIN-2259 Also, opened an issue on tensorflow: https://github.com/tensorflow/tensorflow/issues/8401 On Tue, Mar 14, 2017 at 8:38 AM, Felix Cheung <felixcheun...@hotmail.com<mailto:felixcheun...@hotmail.com>> wrote: This sounds like more an issue with Tensorflow when working in notebook environment in general. Would you please open a JIRA. Also feel free to open an issue on Tensorflow as well. ________________________________ From: Sriram Narayanamoorthy <nsrira...@gmail.com<mailto:nsrira...@gmail.com>> Sent: Tuesday, March 14, 2017 7:28:12 AM To: users@zeppelin.apache.org<mailto:users@zeppelin.apache.org> Subject: Tensorflow not working in Zeppelin 0.7.0 Hi all, There seems to be some issue using Tensorflow in Zeppelin 0.7.0 and it throws this error: NameError: name '_interactive' is not defined Another user reported the same issue<http://stackoverflow.com/questions/42757433/tensorflow-can-not-work-with-zeppelin> on SO. The fix suggested there seems to be really hacky. The issue seems to be that in tf_logging.py file [.\Anaconda3\Lib\site-packages\tensorflow\python\platform\tf_logging.py] the _interactive variable is not triggering for Zeppelin. # Determine whether we are in an interactive environment try: # This is only defined in interactive shells if _sys.ps1: _interactive = True except AttributeError: # Even now, we may be in an interactive shell with `python -i`. _interactive = _sys.flags.interactive Any thoughts why? I am using Zeppelin on Windows. Best, -Sriram -- Sriram Narayanamoorthy -- Sriram Narayanamoorthy -- Sriram Narayanamoorthy
