IPython Notebook kernel dies when importing tensorflow

[shekkizh]

Environment info

Operating System: nvidia-docker (Docker version 1.11.2, build b9f10c9)
Tensorflow version 0.8.0

Steps to reproduce

1. Run ipyhton notebook from nvidia-docker
2. import tensorflow
3. Kernel dies with Kernel Restarting dialog and message "The kernel appears to have died. It will restart automatically"
4. importing numpy package before importing tensorflow solves the issue.

[yaroslavvb]

There was a similar crash fixed in 0.9
#2173 (comment)

[concretevitamin]

@shekkizh can you test against 0.8.0, as @yaroslavvb mentioned (thanks!)?

[shekkizh]

Closing issue since the problem is reported to be fixed in 0.9. Thanks @yaroslavvb @concretevitamin

[tammy228]

i was having the same problem, but i cant solved it by importing numpy package before tensorflow

[er-nabin-bhusal]

i am having same problem in my desktop even when i import numpy before importing tensorflow

[minda163]

I am having same problem in my desktop, 'import numpy before importing tensorflow 'it doesn't work.

[tpaixao]

same problem here, both in python2.7 and python3.5. crash happens even if imported in a normal python shell (no jupyter).

my problem was that my processor was too old. installing tensorflow 1.5.0 solved the problem. Another solution would be to build from source...