This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(294.40000000000003, 60%, 38%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EPR%3C/text%3E%3C/svg%3E)
I created a new compute based on an earlier one from Dec 2023.
Dec 23 compute has Python3.8
New compute has Python3.8 and Python3.9.
Azure notebook uses Python 3.9.19
From command line "import sklearn" works for 3.8 and 3.9
Azure notebook can't import - No module named 'sklearn'
Kernel selected is "Python3.8 Pytorch and Tensorflow" in notebook
Don't have a fix for this. Suggestions?
@Phelps, Ronald (Ron) (CIV) I think you need to install the same using pip magic command i.e using % inside a notebook. Here is a screen shot where the same works after installing using magic command.
Please check if this works for you. Thanks!!
It seems like the issue is that the required libraries are not installed in the Azure Notebook environment. Instead of using !pip install... to install packages in Azure ML notebooks, you can use %pip line magic within a notebook or install from a terminal. Here are the steps to install the sklearn package from a terminal:
conda activate azureml_py38
pip install --upgrade scikit-learn
This should install the sklearn package and allow you to import it in your notebook.
References:
Lots of incompatibilities. See below. This is entirely a conda issue as the Python3.8 and Python3.9 command line interface allows me to import sklearn.
Previous compute instance worked fine and has a scikit-learn version 0.22.1
Requirement already satisfied: numpy>=1.19.5 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (1.23.5)
Requirement already satisfied: joblib>=1.2.0 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (1.2.0)
Requirement already satisfied: threadpoolctl>=3.1.0 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (3.4.0)
Installing collected packages: scipy, scikit-learn
Attempting uninstall: scipy
Found existing installation: scipy 1.5.3
Uninstalling scipy-1.5.3:
Successfully uninstalled scipy-1.5.3
Attempting uninstall: scikit-learn
Found existing installation: scikit-learn 1.1.3
Uninstalling scikit-learn-1.1.3:
Successfully uninstalled scikit-learn-1.1.3
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
skl2onnx 1.14.1 requires scikit-learn<1.3,>=0.19, but you have scikit-learn 1.5.1 which is incompatible.
scikit-image 0.22.0 requires networkx>=2.8, but you have networkx 2.5 which is incompatible.
responsibleai 0.33.0 requires ipykernel<=6.8.0, but you have ipykernel 6.29.4 which is incompatible.
econml 0.15.0 requires scikit-learn<1.5,>=1.0, but you have scikit-learn 1.5.1 which is incompatible.
econml 0.15.0 requires shap<0.44.0,>=0.38.1, but you have shap 0.44.0 which is incompatible.
azureml-training-tabular 1.56.0 requires scikit-learn<=1.1.3,>=1.0.0, but you have scikit-learn 1.5.1 which is incompatible.
azureml-training-tabular 1.56.0 requires scipy<1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.
azureml-train-automl-runtime 1.56.0 requires scikit-learn==1.1.3, but you have scikit-learn 1.5.1 which is incompatible.
azureml-train-automl-runtime 1.56.0 requires scipy<=1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.
azureml-automl-runtime 1.56.0 requires scikit-learn==1.1.3, but you have scikit-learn 1.5.1 which is incompatible.
azureml-automl-runtime 1.56.0 requires scipy<=1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.
Lots of dependency errors. See below
Python3.8 and 3.9 from command line in terminal can import sklearn successfully.
I suspect that the notebook env is the issue.
Requirement already satisfied: numpy>=1.19.5 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (1.23.5)
Requirement already satisfied: joblib>=1.2.0 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (1.2.0)
Requirement already satisfied: threadpoolctl>=3.1.0 in /anaconda/envs/azureml_py38/lib/python3.9/site-packages (from scikit-learn) (3.4.0)
Installing collected packages: scipy, scikit-learn
Attempting uninstall: scipy
Found existing installation: scipy 1.5.3
Uninstalling scipy-1.5.3:
Successfully uninstalled scipy-1.5.3
Attempting uninstall: scikit-learn
Found existing installation: scikit-learn 1.1.3
Uninstalling scikit-learn-1.1.3:
Successfully uninstalled scikit-learn-1.1.3
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
skl2onnx 1.14.1 requires scikit-learn<1.3,>=0.19, but you have scikit-learn 1.5.1 which is incompatible.
scikit-image 0.22.0 requires networkx>=2.8, but you have networkx 2.5 which is incompatible.
responsibleai 0.33.0 requires ipykernel<=6.8.0, but you have ipykernel 6.29.4 which is incompatible.
econml 0.15.0 requires scikit-learn<1.5,>=1.0, but you have scikit-learn 1.5.1 which is incompatible.
econml 0.15.0 requires shap<0.44.0,>=0.38.1, but you have shap 0.44.0 which is incompatible.
azureml-training-tabular 1.56.0 requires scikit-learn<=1.1.3,>=1.0.0, but you have scikit-learn 1.5.1 which is incompatible.
azureml-training-tabular 1.56.0 requires scipy<1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.
azureml-train-automl-runtime 1.56.0 requires scikit-learn==1.1.3, but you have scikit-learn 1.5.1 which is incompatible.
azureml-train-automl-runtime 1.56.0 requires scipy<=1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.
azureml-automl-runtime 1.56.0 requires scikit-learn==1.1.3, but you have scikit-learn 1.5.1 which is incompatible.
azureml-automl-runtime 1.56.0 requires scipy<=1.11.0,>=1.0.0, but you have scipy 1.13.1 which is incompatible.