Azure-docs: Got error - "AttributeError: module 'tensorflow' has no attribute 'logging'" from a pre-configured Notebook VM

[jz247]

Got the following error when I try to run the cell that creates autoMLConfig object from a pre-configured Azure Notebook VM. Any suggestion on how to fix this? Thanks!

AttributeError Traceback (most recent call last)
in
----> 1 from azureml.train.automl import AutoMLConfig
2
3 automl_config = AutoMLConfig(task='regression',
4 debug_log='automated_ml_errors.log',
5 X=x_train.values,

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/train/automl/__init__.py in
23 # Suppress the warnings at the import phase.
24 warnings.simplefilter("ignore")
---> 25 from ._automl import fit_pipeline
26 from .automlconfig import AutoMLConfig
27 from .automl_step import AutoMLStep, AutoMLStepRun

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/train/automl/_automl.py in
17 from automl.client.core.runtime.cache_store import CacheStore
18 from automl.client.core.runtime import logging_utilities as runtime_logging_utilities
---> 19 from azureml.automl.core import data_transformation, fit_pipeline as fit_pipeline_helper
20 from azureml.automl.core.automl_pipeline import AutoMLPipeline
21 from azureml.automl.core.data_context import RawDataContext, TransformedDataContext

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/automl/core/fit_pipeline.py in
18 from automl.client.core.common.limit_function_call_exceptions import TimeoutException
19 from automl.client.core.runtime.datasets import DatasetBase
---> 20 from . import package_utilities, pipeline_run_helper, training_utilities
21 from .automl_base_settings import AutoMLBaseSettings
22 from .automl_pipeline import AutoMLPipeline

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/automl/core/pipeline_run_helper.py in
18 from automl.client.core.common.exceptions import ClientException
19 from automl.client.core.runtime import metrics
---> 20 from automl.client.core.runtime import pipeline_spec as pipeline_spec_module
21 from automl.client.core.runtime.datasets import DatasetBase
22 from automl.client.core.runtime.execution_context import ExecutionContext

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/automl/core/_vendor/automl/client/core/runtime/pipeline_spec.py in
21
22 from automl.client.core.common import constants
---> 23 from automl.client.core.runtime import model_wrappers, tf_wrappers
24 from automl.client.core.runtime.nimbus_wrappers import AveragedPerceptronBinaryClassifier, \
25 AveragedPerceptronMulticlassClassifier, NimbusMlClassifierMixin, NimbusMlRegressorMixin

/anaconda/envs/azureml_py36/lib/python3.6/site-packages/azureml/automl/core/_vendor/automl/client/core/runtime/tf_wrappers.py in
34 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
35 if tf_found:
---> 36 tf.logging.set_verbosity(tf.logging.ERROR)
37
38 OPTIMIZERS = {

AttributeError: module 'tensorflow' has no attribute 'logging'

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: 68a3df81-6871-c443-c413-16bb97982bd4
• Version Independent ID: 5030fc51-ec28-d96e-fe85-a06a8fa1ce6f
• Content: Regression model tutorial: Automated ML - Azure Machine Learning
• Content Source: articles/machine-learning/service/tutorial-auto-train-models.md
• Service: machine-learning
• Sub-service: core
• GitHub Login: @trevorbye
• Microsoft Alias: trbye

[RohitMungi-MSFT]

@jz247 Thanks for the feedback. We are investigating the issue and will update you shortly.

[RohitMungi-MSFT]

@jz247 It looks like the pre-reqs to install some of the pip packages might be missing in this environment. Could you please try to install the same?
!pip install azureml-sdk[automl] azureml-opendatasets azureml-widgets

Also, please check which version of tensorflow is installed

import tensorflow as tf 
print(tf.__version__)

It also looks like a higher version(>2) of tensorflow is installed which might not work for this tutorial. Could you try downgrading tensorflow version and re-run the tutorial?

[jz247]

Thank you so much for your prompt reply & info! Sure - will try it soon.

[jz247]

@RohitMungi-MSFT, run the install line through JupyterLab Terminal. Checked the TensorFlow version (=2.0.0). Still got the same error as before. Any other things that I can try? Thanks!

FYI - The Azure Notebook VM that I am using was created on Oct. 20.

[RohitMungi-MSFT]

@jz247 Please downgrade the tensorflow version. Please try to use version 1.12.0 and try the same.

[jz247]

Thanks for your prompt reply, @RohitMungi-MSFT! Will do.

[jz247]

It works. Thank you so much for your help, @RohitMungi-MSFT!

[RohitMungi-MSFT]

@jz247 We will now proceed to close this thread. If there are further questions regarding this matter, please tag @RohitMungi-MSFT in your reply. We will gladly continue the discussion and we will reopen the issue.

[RobotChipmunk]

@RohitMungi-MSFT I got this same error when trying to run the autoML sample which is part of the default Azure notebook samples. I was running the notebook on a notebook VM created in Azure. The sample file i am using is called 'regression-automated-ml'

[Ajithvajrala23]

I am having the same issue. But I couldn't downgrade tensorflow to 1.12.0 on Jupyter Notebook as mentioned above in one of the comments, I tried downgrading on Azure Work-space.
Is this issue resolved yet ?
Is there any way on Azure work-space where user has control over package versions provided by Azure? (I mean upgrading or downgrading versions based on user preferences like in Pycharm?)

[yuukiiwa]

Hi @Ajithvajrala23 and @RobotChipmunk,
I got this problem too but have figured out how to solve them, and here are the steps:

1. In your virtual machine learning Jupyter notebook, click on "New" button on the upper left corner and get a new terminal.
2. move all your files from User to the directory with your Azure name (mine is wanyuk), so for me, I will type this in my User directory

mv file.ipynb [wanyuk]

1. get into the directory with your Azure name

cd [wanyuk]

1. downgrade Tensorflow using pip

pip install tensorflow==1.12.0

1. get into your Jupyter notebook .ipynb and check the version with

import tensorflow as tf 
print(tf.__version__)        

Hope this helps.

Best,
Yuk Kei