Ambiguous error in Azure Machine Learning Designer 'Evaluate Model' Module

Question

Ambiguous error in Azure Machine Learning Designer 'Evaluate Model' Module

Sampat, Varun 236

I am getting the following error from the Evaluate Model module in Azure Machine Learning Designer:

When I open the Assigned Data to Clusters module everything seems fine. I downloaded the output for Assigned Data to Clusters and played with cluster number 31 and there doesn't seem to be any issue. Additionally, I am using Azure Modules, so I am confused as to why this is failing. Please provide some clarity into this issue. This is a part of my pipeline:

Additionally, it seems unless I successfully run the Evaluate Model module, I cannot create an inference pipeline. If this is untrue, please help me out here as well. There is no option for me to 'Create an Inference Pipeline' which shown in this tutorial; step 1.

Please let me know if you need any other information.

Thanks in advance.

GiftA-MSFT 11,176 Reputation points

2021-06-01T22:06:02.287+00:00
Hi, you should be able to deploy your clustering model without using Evaluate Model module. Regarding the error message, can you share additional error details? You can find the error message in the designer following these steps:

Select the failed module, go to the Outputs+logs tab, you can find the detailed log in the 70_driver_log.txt file under the azureml-logs category.

For detailed module error, you can check it in the error_info.json under module_statistics category.
Sampat, Varun 236 Reputation points

2021-06-02T12:34:43.34+00:00
Hi,

How can I generate an inference pipeline automatically? This is what I see:

But this is what Azure docs show I should be able to see:

According to this tutorial, I need to be able to create an inference pipeline before moving forward.

I have also attached the log files you have requested; I am removing potential sensitive data. Hope that helps:

101698-error-info.txt

101699-70-driver-log.txt

Thank you in advance!
GiftA-MSFT 11,176 Reputation points

2021-06-02T16:28:35.357+00:00

You need to first submit to train the model, then the option to deploy inference pipeline will become available. You can try to submit without the Evaluate Model module, then after training, you should see create inference pipeline option. Regarding the error, I'm trying to gather more information internally and will share details accordingly.

Accepted answer

0 additional answers

Your answer

GiftA-MSFT 11,176 Reputation points

2021-06-01T22:06:02.287+00:00

Hi, you should be able to deploy your clustering model without using Evaluate Model module. Regarding the error message, can you share additional error details? You can find the error message in the designer following these steps:

Select the failed module, go to the Outputs+logs tab, you can find the detailed log in the 70_driver_log.txt file under the azureml-logs category.

For detailed module error, you can check it in the error_info.json under module_statistics category.
Sampat, Varun 236 Reputation points

2021-06-02T12:34:43.34+00:00

Hi,

How can I generate an inference pipeline automatically? This is what I see:

But this is what Azure docs show I should be able to see:

According to this tutorial, I need to be able to create an inference pipeline before moving forward.

I have also attached the log files you have requested; I am removing potential sensitive data. Hope that helps:

101698-error-info.txt

101699-70-driver-log.txt

Thank you in advance!
GiftA-MSFT 11,176 Reputation points

2021-06-02T16:28:35.357+00:00

You need to first submit to train the model, then the option to deploy inference pipeline will become available. You can try to submit without the Evaluate Model module, then after training, you should see create inference pipeline option. Regarding the error, I'm trying to gather more information internally and will share details accordingly.

Answer 1

Can you please check if the Assignment cluster 31 has NaN value? The Assign Data to Clusters leverages SKlearn, and from the error message, seems the Assignment column had NaN value which resulted in an error. If that's the case, let us know, so we can enable Evaluate Module module to deal with NaN values, and in the meantime, here's a short-term workaround:

Connect Clean Missing Data module to Assign Data to Cluster module, to clean the missing values.
Use Edit Metadata module to convert Assignment to Integer and categorical type, this is because if Assignment column has NaN value before and its column type was double, we need to convert it to integer.
Connect Edit Metadata to Evaluate Model module.

Hope this help!

Share via

Ambiguous error in Azure Machine Learning Designer 'Evaluate Model' Module

0 additional answers

Your answer