Python module error while running KERAS Network Learner node

[nilotpalc]

Hello all,
I am trying to run the workflow for the example “Glucose TSA” available in the Codeless Time Series Analysis with KNIME space in KAP 5.1.1 on my laptop.

knime/Codeless Time Series Analysis with KNIME – Chapter 9

Running the workflow fails the Network Learner Node with the following message -

Traceback (most recent call last):
File “”, line 2, in
File “C:\Program Files\KNIME\plugins\org.knime.dl.python_5.1.0.v202303061416\py\DLPythonNetwork.py”, line 53, in
from pandas.util.testing import network
ModuleNotFoundError: No module named ‘pandas.util.testing’

I checked the py file in the specified location. There is statement in the file as below
from pandas.util.testing import network

I checked ChatGPT and it advised to remove this statement. I ran the code file post removing this statement on my machine and in the same environment. I did not get any error.

The import from pandas.util.testing import network is problematic as pandas.util.testing was deprecated and later removed in newer versions of pandas. Since there’s no reference to network in the rest of the code, this line can be safely removed.

Can I please request if the KNIME team can check the same and help to resolve the error?

Any help and steer is highly appreciated.

–
Best Regards,
nilotpalc

[mlauber71]

@nilotpalc as a quick hint you will have to set up the necessary Python packages in versions compatible with the current knime nodes

Deep Learning with Keras and TensorFlow -- Troubleshooting

@rrembert3 I think the current implementation of Keras and Tensorflow of KNIME would maybe need different settings. The setup can be tricky - you can extract what combination KNIME recommends and try to recreate that the best you can. My current solution is to use different Python environments for Keras and Tensorflow. But I have not worked on that for some time. So you will have to explore. I have some examples here:

AttributeError: module ‘keras.optimizers’ has no attribute ‘Adam’

@akhtarameen as it happens I just have collected some example how to set up Keras and Tensor Flow 2 (for classification tasks) with the help of the Conda Environment propagation. If you have a working Miniconda or Anaconda environment configured in KNIME you should be able to just use the propagation and get a working deep learning setup. You would then have to carefully check and plan what you want to d with the setup and what architecture you want to use. [image] [image] The setup in …

https://docs.knime.com/latest/deep_learning_installation_guide/index.html

[nilotpalc]

Thanks @mlauber71
I had created a separate environment in conda specific to KNIME DL.
Let me follow the steps and use the ‘new environment’ within Preferences to create a new environment and see if the error goes away.

–
Best Regards,
nilotpalc

[nilotpalc]

Hello @mlauber71
I created new environment for KERAS following the instructions in the guide.
I am now getting the below mentioned error.

I am not a pro-python user and hence, unable to understand the errors indicating a metaclass conflict.

Is this possibly linked to my other environments?

–
Best Regards,
nilotpalc

[mlauber71]

@nilotpalc setting up DL environment sis not easy unfortunately. You can do these things: Set up conda and just use these two metanodes:

AutoML - All Models Post Process Component Failing

It should work on windows. I also have two meta nodes with Conda Environment Propagation for Keras and TensorFlow for Windows. I have not tested them for some time though.

Or if this fails read this article:

KNIME and Python — Setting up Deep Learning Environments for Keras and...

[nilotpalc]

Hi @mlauber71
Thank you for sharing the above links.
I primarily decided to use the component for Keras environment creation.

knime_py_keras_win – mlauber71

It was able to successfully create the environment and there were no errors in running the workflow specific to the Keras Learner Node.

I also created a yml file basis the annotation inside the component and created a fresh environment. I configured the new environment in the KNIME preferences for Keras .The newly created environment also worked without any errors.

For new users visiting this post, please make sure to change the configuration in the Learner Node in case you are switching between the configuration component and Knime preferences.

nilotpalc/Public – Keras YML File

Other key takeaway from this experience is the creation of the environment through the Preferences → Python Deep Learning → New Environment method as elaborated in the Installation guide works fine for Tensor Flow 2 as shared above.

@mlauber71 Once again, thanks for all the guidance.

–
Best Regards,
nilotpalc