Getting water.exception for H2O Partial Plot

Question

I am getting H2OResponseError: Server error water.exceptions.H2OIllegalArgumentException: Error: unimplemented

H2OResponseError: Server error water.exceptions.H2OIllegalArgumentException:
  Error: unimplemented
  Request: POST /3/PartialDependence/
    data: {'cols': '[offer_id]', 'model_id': 'GBM_grid__1_AutoML_20200510_220603_model_14', 'frame_id': 'digiq_wine_multiclass_training_1__scrubbed3.hex', 'nbins': '20', 'add_missing_na': 'False', 'row_index': '-1'}

When I try to find PDP for one of the datasets I am working with. This doesn't break for all the dataset but, in few peculiar cases like this one, it does. I am not able to find the reason, and not sure if this could be a bug of H2O. Having checked the log file, it shows the same error as above with no other relevant information. H2O_cluster_version: 3.30.0.1 Language: Python 3.7.6

I am providing the reproducible example pointing to my GitHub to readily see the visible error also can be downloaded and just run to experience it. I would appreciate it if anyone from the H2O team or relevant person to help in resolving or find the root cause of the issue. Thank you.

GitHub replicating example - https://github.com/prabhuSub/H2O-Water-Exception

Notebook in the repo - https://github.com/prabhuSub/H2O-Water-Exception/blob/master/Sample_issue.ipynb

logs - https://github.com/prabhuSub/H2O-Water-Exception/tree/master/h2ologs_node0_127.0.0.1_54321

The log file is added too for reference if required, for a quick investigation.

Erin LeDell Erin LeDell · Accepted Answer · 2020-05-13T04:57:56

Your dataset is called "digiq_wine_multiclass_training(1)_scrubbed.csv" so I assume you were have a multiclass classification problem? If so, then the reason you're hitting an error is because partial dependence plots are only implemented for regression or binary classification problems.

Getting water.exception for H2O Partial Plot

1 Answers