AttributeError: /anaconda/envs/azureml_py36/lib/libxgboost.so: undefined symbol: XGBoosterUnserializeFromBuffer

Question

I am following automated ML guideline.:

https://github.com/Azure/MachineLearningNotebooks/blob/master/how-to-use-azureml/automated-machine-learning/regression/auto-ml-regression.ipynb

I created development environment according to the documentation. As recommended, I am using version 1.27.0 of the Azure ML SDK.

After remote_run completed, I got the error for the following command "best_run, fitted_model = remote_run.get_output()"

I would appreciate if you could help me.

Answer 1

@Cagatay Topcu Thanks for the question. Can you please try manually downgrade xgboost with: pip install xgboost==0.90 or Create a new Compute Instance which should have xgboost==0.90.