Model evaluation job issue

[txn]

I am trying to create a model evaluation in Vertex AI using the Create Evaluation option. Per the documentation (https://cloud.google.com/vertex-ai/docs/evaluation/using-model-evaluation), "Vertex AI automatically sends an email notification when a model evaluation job is complete." 

It's been almost two days and I still haven't received the email notification, and the UI still says 'No model evaluations created yet' Screenshot. Any ideas to help me troubleshoot this? 

I am using a test data from BigQuery with 1000 rows to evaluate the model.

[Poala_Tenorio]

First, ensure that the evaluation job was actually initiated and is running. You can check the status of your evaluation job in the Vertex AI Console. If it's still in progress, wait for it to complete.

Double-check the configuration of your evaluation job to ensure that you provided the correct inputs, including the model name, dataset, and other parameters. Any misconfiguration can lead to issues in the evaluation process.

Inspect the logs and error messages associated with your evaluation job. Look for any error messages or issues that might have occurred during the evaluation process. Address any errors or problems found.

Ensure that the BigQuery dataset you are using for evaluation is accessible and contains the expected data. Make sure the dataset is correctly linked to your evaluation job.

Check if your evaluation job is using excessive resources or is stuck in a resource-intensive task. This could lead to delays in completion. Ensure that you have sufficient resources available. If you suspect there was an issue with the initial evaluation job, consider retrying it with the same or slightly modified parameters to see if it completes successfully this time.

Remember that email notifications are sent when the evaluation job is completed successfully. If you've followed the steps above and still haven't received any notifications after a reasonable amount of time, it's essential to investigate any potential issues in the evaluation job itself and ensure that it's running correctly and has access to the required resources and data. If the problem persists and you have ruled out configuration and resource issues, consider reaching out to Google Cloud Support for assistance. They can help investigate and resolve platform-specific problems.

[txn]

In the XGBoost model evaluation logs I get the below error (screenshot: No valid prediction data found in batch_prediction_gcs_source directory). But when I look at the batch prediction logs stored in GCS for the same job, I see an error related to json data format (“error”: “Prediction failed: Could not initialize DMatrix from inputs: (\‘Expecting 2 dimensional numpy.ndarray, got: \’, (2,))“).How should be the data be formatted for the evaluation of XGBoost model?I tried using formats like these:

1. {“instances”: [[0,1,2,….],[10,11,12,….]], “rank”: [10,15]}
2. {“instance”: [0,1,2,….],, “rank”: 10}

{“instance”: [10,11,12,….],, “rank”: 15}
where “rank” is the target variable both of these formats gave the same error