Attributeerror: module ‘tensorflow’ has no attribute configproto error occurs configproto is completely removed in TensorFlow 2.0. This configproto module does not provide back word compatibility in TensorFlow 2.0. There are multiple ways to fix this AttributeError. We will explore them in this article.
Attributeerror: module ‘tensorflow’ has no attribute configproto ( Solution ) –
Solution 1: Using Compatible Syntax –
If you want to use any syntax from TensorFlow 1.x series in TensorFlow 2.x then you can use the below way to import the statement.
in the place of tf.ConfigProto . It will only make this import compatible but if you want to run complete 1.x code in 2.x interpreter. Then you should add two lines to your code base.
import tensorflow.compat.v1 as tf tf.disable_v2_behavior()
Solution 2: Converting the code from TensorFlow 1.0 to 2.x –
The other workaround to fix this error is migrating the code base from 1.x to 2.x if the underline interpreter is 2.0-supportive. Here is the command for this migration.
tf_upgrade_v2 \ --intree my_project_1.x/ \ --outtree my_project_v2/ \ --reportfile report.txt
Here my_project_1.x is on tensorFlow 1.x and we are migrating it into my_project_v2 . In between the report.txt works as log for this migration.
Solution 3: Downgrading TensorFlow to 1.x series –
Since the code is breaking because of version incompatibility. Then one way is to downgrade TensorFlow version to the 1.x version but this is not recommended because it may break more worse if another part of the code also does not provide backward compatibility.
pip install tensorflow==1.15.5
You may try other lower versions as well. but make sure it is not breaking further.
Similar Errors –
Since there are multiple breaking changes in Tensorflow Migration Release 1.x to 2.x. Due to this, the Interpreter throws similar errors.
Data Science Learner Team
Join our list
Subscribe to our mailing list and get interesting stuff and updates to your email inbox.