Remote run model unable to be saved

Bernardo Jaccoud 21

I've built models using the AutoML function and I'm trying to call the best model to deploy into production. The AutoML function ran correctly and produced the ~35 models. My goal is to pull the best model. Here is the code:

best_run, fitted_model = remote_run.get_output()
fitted_model

When runnning the code, I get the following error:

AttributeError: 'DataTransformer' object has no attribute 'enable_dnn'

Any help would be much appreciated.

Accepted answer

romungi-MSFT 47,441 Reputation points Microsoft Employee

2021-02-12T11:23:05.603+00:00

@Bernardo Jaccoud Did your run configure enable_dnn i.e bert settings of automated ML? I am curious to understand what the status of your run is directly on the portal ml.azure.com?
Please sign in to rate this answer.
Bernardo Jaccoud 21 Reputation points

2021-02-12T16:54:29.073+00:00

Hey, thanks for the knowledge! The combination of updating the cluster to Standard NC6 and setting 'enable_dnn' to False worked. Based on the fact that enable_dnn default = False, it seems as if changing the cluster did the trick. It seems as if BERT is only enabled using the Standard NC6 cluster.

Am I thinking about this right?

romungi-MSFT 47,441 Reputation points Microsoft Employee

2021-02-15T05:24:10.117+00:00

@Bernardo Jaccoud Yes, the compute needs to be Standard NC6 or higher type in this case. Glad, that this worked for you. It would be great if you could accept the same answer for better visibility to others for similar issues.
Sign in to comment

Use comments to ask for clarification, additional information, or improvements to the question.

Your answer

Share via

Remote run model unable to be saved

0 additional answers

Your answer