Azure Machine Learning Designer: Cannot create inference pipeline because there is no trained model in this pipeline - When using cross validation

Question

Azure Machine Learning Designer: Cannot create inference pipeline because there is no trained model in this pipeline - When using cross validation

Gustavo Avino 0

I can't create an inference pipeline if I use the cross validate option instead the classic split/train model.
Is it because to create an inference pipeline you must have the "train model" component? If so, how can you use the cross validate model and why? because at the end you have to use the "train model" to deploy.

1 answer

Your answer

Answer 1

YutongTie-MSFT 53,986 Moderator

Hello @Gustavo Avino

Thanks for reaching out to us, you need to submit your pipeline first, after submitting, you can go to your job details then you should see create inference pipeline button.

I have some issue when I first touched the new UI, I can not see the button because I zoom my page in 100%, please try 50% to see if you can not see the button.

Please let me know if you still have the issue, I am happy to help further, please kindly share a screenshot of your pipeline in that situation for us to reproduce if that is not confidential.

Regards,

Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.

Gustavo Avino 0 Reputation points

2023-06-30T08:06:23.61+00:00

Thanks for the answer!. The problem is once i have submitted my pipeline, I cannot create an inference pipeline. It only happens if I use cross validate model, not when I use train/split.
Gustavo Avino 0 Reputation points

2023-06-30T15:17:12.5833333+00:00
YutongTie-MSFT 53,986 Reputation points Moderator

2023-07-03T16:00:32.1633333+00:00

Hello @Gustavo Avino

Thanks for sharing the screenshot, I have brought this scenario to product team and waiting for the response. Will let you know once we get any news.

Regards,

Yutong
YutongTie-MSFT 53,986 Reputation points Moderator

2023-07-11T03:45:17.4033333+00:00

Hello @Gustavo Avino

Thanks for your waiting, I just got a confirmation from product team as below -

This is by design behavior. To create a inference pipeline in designer, the pipeline graph needs to only generate one trained model. The cross validation will train multiple models depending on the number of folds. So, the system cannot recognize know which model to use in the inference pipeline. To unblock, please use Train model module, as described in Tutorial: Designer - train a no-code regression model - Azure Machine Learning | Microsoft Learn

I hope this helps, please kindly accept the answer if you feel helpful to support the community, thanks a lot.

Regards,

Yutong

Share via

Azure Machine Learning Designer: Cannot create inference pipeline because there is no trained model in this pipeline - When using cross validation

1 answer

Your answer