I am using qiime2 2019.4 in conda. I was going over the moving pictures tutorials for the taxonomy. I ran the following commands:
qiime feature-classifier classify-sklearn \
The error message I received: The scikit-learn version (0.21.2) used to generate this artifact does not match the current version of scikit-learn installed (0.21.3). Please retrain your classifier for your current deployment to prevent data-corruption errors.
I can’t figure out how to retrain the greengenes 99% that is used in this tutorial with the new version of scikit-learn…I have seen similar topics in the forum but the opposite case, where the classifier was trained for scikit-learn 20.2 and the person had installed a version 19, so they only had to update the version. I can’t figure out how to retrain the classifier using the newer version of scikit-learn 20.3
thank you
Have you seen the QIIME 2 tutorials? There is one about training a feature classifier.
That is your other option: to downgrade scikit-learn to the correct version. Something like this should work from within your activated QIIME 2 environment (but see the conda documentation if this does not work):
conda install scikit-learn=0.21.2
then confirm that you have the desired version installed with this command:
I tried to reinstall the scikit-learn version that works with the tutorial…it didn’t work, it is still said that the scikit-learn version is 0.21.3. I went over the tutorial to retrain the classifier using the dataset from the tutorial. When I run the script to train the classifier it returns the following error: Plugin error from feature-classifier:
fit_classifier_naive_bayes() got an unexpected keyword argument 'feat_ext__non_negative’
Debug info has been saved to /var/folders/8s/1_k7c79906dglglh6_91yp9r0000gp/T/qiime2-q2cli-err-o8pljva4.log