Azure ML designer - Error in scoring W&D Recommender when using All items or Unrated items

Question

Starting from the example pre-build of the Restaurant Ratings (Wide and Deep Recommender) pipeline:

1. I trained the model using the score setting "From rated items (for model evaluation)
2. When job finished succesfully I created an online-inference pipeline as shown below (auto-created + added web serivce input that the designer "forgot" to add automatically 3.
3. Set the scoring setting of the component to "From unrated items" to suggest items to "new users", following the documentation I had to connect to third port of the component the ratings dataset to "filter" out the knows users already trained
4. saved and submitted his pipeline I get the error:

azureml.studio.internal.error.LibraryExceptionError: Unknown library exception: type object 'object' has no attribute 'dtype'. Please contact product team by submit a feedback. You can submit a feedback by clicking the face icon on top right corner of the page. .

ModuleExceptionMessage:Library Error - AttributeError: type object 'object' has no attribute 'dtype'

The above was not happening in December.

I've also try to get the model in datastore and run the score.py in a Jupyter notebook getting the same error.

some libraries are not correctly configured in the Studio I believe, any help on the above?

Thanks,

AGA

std_log.txt

Answer 1

Hi guys! I'm having the same issue with a recommender that i'm building. When changing the parameter from "From rated Items" to "All items" that same error appears. Alberto G [A*] Ramr-msft did you happen to resolve it some how?