Discover all of the brand-new features and improvements to existing capabilities in the Dataiku 11.3 updateLET'S GO

ModuleNotFoundError: No module named 'sklearn.inspection'

Khalid538
Level 2
ModuleNotFoundError: No module named 'sklearn.inspection'

One of our SME is trying to build a ML model using the package scikit-learn and this package is already present in our default Python environment but she is getting following error in the Jupyter Notebook. i have attached the error code.


Operating system used: Cent OS

0 Kudos
2 Replies
CatalinaS
Dataiker

Hi @Khalid538

The error indicates that the package scikit-learn is missing from your default Python environment. As stated in the documentation installing Python packages in the built-in DSS environment is not recommended. In addition trying to install the package using pip directly in the notebook is not advised. The solution is to use code environments to create a code environment and then add scikit-learn to the list of required packages. After that, you can change the code environment in the notebook by going to Kernel>Change Kernel and select the code environment that contains the needed package.

0 Kudos
AlexT
Dataiker

Just to Catalina's response. The reason you should create a custom code env is sklearn.inspection was introduced in version 0.21.0. The default DSS version is 0.20.4. Hence the error you are receiving the "No module named" error. 

To use sklearn.inspection you need to create a custom code env with the requirement:

scikit-learn>=1.0,<1.1

Note there is experimental support DSS 11 there experimental support for scikit-learn 1. This version should work with visual ML. As well if you don't need Visual ML then you are free to use another version > 0.2.1

Hope this helps.

0 Kudos