[tensorflow] Module 'tensorflow' has no attribute 'contrib'

I am trying to train my own custom object detector using Tensorflow Object-Detection-API

I installed the tensorflow using "pip install tensorflow" in my google compute engine. Then I followed all the instructions on this site: https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html

When I try to use train.py I am getting this error message:

Traceback (most recent call last): File "train.py", line 49, in from object_detection.builders import dataset_builder File "/usr/local/lib/python3.6/dist-packages/object_detection-0.1->py3.6.egg/object_detection/builders/dataset_builder.py", line 27, in from object_detection.data_decoders import tf_example_decoder File "/usr/local/lib/python3.6/dist-packages/object_detection-0.1-py3.6.egg/object_detection/data_decoders/tf_example_decoder.py", line 27, in slim_example_decoder = tf.contrib.slim.tfexample_decoder AttributeError: module 'tensorflow' has no attribute 'contrib'

Also I am getting different results when I try to learn version of tensorflow.

python3 -c 'import tensorflow as tf; print(tf.version)' : 2.0.0-dev20190422

and when I use

pip3 show tensorflow:

Name: tensorflow Version: 1.13.1 Summary: TensorFlow is an open source machine learning framework for everyone. Home-page: https://www.tensorflow.org/ Author: Google Inc. Author-email: [email protected] License: Apache 2.0 Location: /usr/local/lib/python3.6/dist-packages Requires: gast, astor, absl-py, tensorflow-estimator, keras-preprocessing, grpcio, six, keras-applications, wheel, numpy, tensorboard, protobuf, termcolor Required-by:

    sudo python3 train.py --logtostderr --train_dir=training/ -- 
    pipeline_config_path=training/ssd_inception_v2_coco.config

What should I do to solve this problem? I couldn't find anything about this error message except this: tensorflow 'module' object has no attribute 'contrib'

If you want to use tf.contrib, you need to now copy and paste the source code from github into your script/notebook. It's annoying and doesn't always work. But that's the only workaround I've found. For example, if you wanted to use tf.contrib.opt.AdamWOptimizer, you have to copy and paste from here. https://github.com/tensorflow/tensorflow/blob/590d6eef7e91a6a7392c8ffffb7b58f2e0c8bc6b/tensorflow/contrib/opt/python/training/weight_decay_optimizers.py#L32

I used google colab to run my models and everything was perfect untill i used inline tesorboard. With tensorboard inline, I had the same issue of "Module 'tensorflow' has no attribute 'contrib'".

It was able to run training when rebuild and reinstall the model using setup.py(research folder) after initialising tensorboard.

I used tensorflow 1.8 to train my model and there is no problem for now. Tensorflow 2.0 alpha is not suitable with object detection API

This issue might be helpful for you, it explains how to achieve TPUStrategy, a popular functionality of tf.contrib in TF<2.0.

So, in TF 1.X you could do the following:

resolver = tf.contrib.cluster_resolver.TPUClusterResolver('grpc://' + os.environ['COLAB_TPU_ADDR'])
tf.contrib.distribute.initialize_tpu_system(resolver)
strategy = tf.contrib.distribute.TPUStrategy(resolver)

And in TF>2.0, where tf.contrib is deprecated, you achieve the same by:

tf.config.experimental_connect_to_host('grpc://' + os.environ['COLAB_TPU_ADDR'])
resolver = tf.distribute.cluster_resolver.TPUClusterResolver('grpc://' + os.environ['COLAB_TPU_ADDR'])
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)