search

Azure-Samples / Custom-vision-service-iot-edge-raspberry-pi

Sample showing how to deploy a AI model from the Custom Vision service to a Raspberry Pi 3 device using Azure IoT Edge
MIT License
186 stars 135 forks source link

image-classifier-module fails on AMD64 image - module 'tensorflow' has no attribute 'GraphDef' #50

Closed nanowireUK closed 4 years ago

nanowireUK commented 5 years ago

Please provide us with the following information:

This issue is for a: (mark with an x)

- [x] bug report -> please search issues before submitting
- [ ] feature request
- [ ] documentation issue or request
- [ ] regression (a behavior that used to work and stopped in a new release)

Minimal steps to reproduce

Clone repo Change deployment.template.json to AMD64 version Build and push IoT solution Let containers run and see image-classifier-service fails iotedge logs image-classifier-service

Any log messages given by the failure

Traceback (most recent call last): File "app.py", line 15, in from predict import initialize, predict_image, predict_url File "/app/predict.py", line 22, in graph_def = tf.GraphDef() AttributeError: module 'tensorflow' has no attribute 'GraphDef'

Expected/desired behavior

It works with included Dockerfile

OS and Version?

Ubuntu 18.04 and Windows

Versions

Mention any other details that might be useful

Can be fixed by changing the Dockerfile to use a different base image and then install fixed tensorflow 1.12.0 and then including the various apt install for python dependencies.

FROM tensorflow/tensorflow:latest-py3

FROM ubuntu:xenial

Thanks! We'll be in touch soon.

nielsams commented 5 years ago

@nanowireUK I had the same issue and was able to do a quick fix by adding 'tensorflow==1.12.0' to the end of the amd64-requirements.txt file and rebuilding.

Awaiting a proper fix of course, but maybe this will help you forward for now

emmanuel-bv commented 4 years ago

Fixed with https://github.com/Azure-Samples/Custom-vision-service-iot-edge-raspberry-pi/pull/62