Vertex AI Auto ML regression training pipeline error message: Internal error occurred

[jalumbau]

First time user of the Vertex AI Auto ML training pipeline. I've made multiple attempts at training a fairly simple dataset (less than 10K observations, no missing data) using tabular regression, and it's always immediately failing with the error "Training pipeline failed with error message: Internal error occurred. Please retry in a few minutes." I've used both MAE and RMSE. Details of one training session below. What am I doing wrong?

StatusFailed

Training pipeline ID5256146067350618112

CreatedJan 7, 2024, 2:44:19 PM

Start timeJan 7, 2024, 2:44:43 PM

Budget (original)1 node hours

Elapsed time15 sec

Regionus-west1

Encryption typeGoogle-managed

Datasetuntitled_1704575609483

Target columnPE

Data splitRandomly assigned (80/10/10)

Column metadataVIEW DETAILS

AlgorithmAutoML

ObjectiveTabular regression

Optimized forMAE

[daviderinayo]

This is also happening to me. And i am so confused because the dataset is very simple.
The logs aren't very helpful either.

The most informative log i found was: "The DAG failed because some tasks failed. The failed tasks are: [tabular-stats-and-example-gen]."
Then the logs for tabular-stats-and-example-gen say 

error.code: 13
error.message: INTERNAL

[daviderinayo]

have you been able to successfully train the model?

[Future]

i Have same propblem here, already wasted some hours to check dataset etc and it seems its just pipeline problem not mine 😞

[davireis]

I am getting the same in my recently created project, but I have an older project where everything just works, even with the same data.

[jalumbau]

No, I haven't trained it yet - that's the step where it's failing. I've read elsewhere that AutoML may have issues with Excel CSV files, which is what I uploaded. I'm going to try saving it as a Google CSV and try that.

[Future]

Thanks for tip! I used Google sheets CSV, but I'll try today than to save
them with Python, I'll let you know if that helped

[daviderinayo]

For me, it was fixed by granting the requested roles to the service account used when creating a pipeline run for training the model.

[mo0o]

I've done granted these roles but still having the error

[mo0o]

I am having the same issue here too