Ladislav Zitka

I am having now same issue: tensorboard==1.13.1 tensorflow==1.13.1 tensorflow-estimator==1.13.0 tensorflow-gpu==1.13.1

Related to #10

Related to #9 - all these are having same root cause seems to me.

To reproduce such issue I did "pipreqs ." to get just project requirments and this is result: ``` six==1.11.0 requests==2.21.0 Flask==1.0.2 numpy==1.16.1 tensorflow_gpu==1.13.1 Pillow==6.0.0 scikit_learn==0.21.1 tensorflow==1.13.1 typing==3.6.6 opecv is provided...

So I run docker now and examine the environment, so we have there following: Python 3.5.2 vs local 3.6.5 Packages; numpy==1.14.2 Flask==0.12.2 opencv_python==3.4.0.12 tensorflow==1.7.0 requests==2.18.4 six==1.11.0 Pillow==6.0.0 scikit_learn==0.21.1 typing==3.6.6

So when I reverted locally all versions via pip to what I find in docker container loaded in docker hub: I get error on startup, but the application starts ok:...

Same error here running from local environment on Gentoo, found some refs: https://ncsforum.movidius.com/discussion/589/keyerror-the-name-input-0-refers-to-a-tensor-which-does-not-exist-when-compile-retrained-model/p2 https://ncsforum.movidius.com/discussion/640/keyerror-the-name-input-0-refers-to-a-tensor-which-does-not-exist I think it runs fine under virutal env, or was at least on my other computer(laptop)...

Mostly I found this, its related to shapes unnamed... need to discover more: https://github.com/Microsoft/vs-tools-for-ai/issues/16

There occured also another error, I will try the docker image from dockerhub first.

Ok, the dockerhum image works fine