Tensorflow 2 API: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead

2
votes

I'm newbie in Tensorflow and I'm developing a notebook in Google Colaboratory.

On my first notebooks I got a message saying to update to Tensorflow v.2.

Now, in another notebook, I get the following message:

The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.

If I do that, am I using Tensorflow v2?
I'm confuse because of the v1 in tf.compat.v1.get_default_graph.

All the error messages are:

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.

WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4267: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.

I think I can't change that code on keras.

2
pythontensorflowtensorflow2.0

2 Answers

3
votes

You still get some of these errors even if you use TF 1.1X, where X = 3/4/5.

Some of the libraries are already being deprecated, although available in TF 1.13/1.14/1.15. Note that TF 1.15 is the latest release for the version 1.X.

To ensure that you are using TensorFlow 2.0, use:

import tensorflow as tf
print(tf.__version__)

There are a couple of notable differences between the TensorFlow 1.X API and TensorFlow 2.X API; since you are a beginner, I would strongly recommend that you stick to the latest versions and take into consideration every deprecation warning, so as to use the newest updates.

The function tf.compat.v1.get_default_graph() is part of the TensorFlow 2.0 API: https://www.tensorflow.org/api_docs/python/tf/Graph?version=stable

2
votes

It’s normal. As you mentioned in the question this warning returns to Keras. Keras is a high level api. It works with theano and tensorflow as backend for low level calculations. If Keras found any version of tensorflow or theano specific code executes that including this part:

tf.get_default_graph

In other side tensorflow doesn’t know who calls it. It throws a warning in future version tf.get_default_graph would be removed! Please be futuristic and don’t use it for code consistency. It returns to Keras not your code. Keras published before tf v2.