How to recover training data using Vertex AI Metadata?

You are investigating the root cause of a misclassification error made by one of your models. You used Vertex AI Pipelines to train and deploy the model. The pipeline reads data from BigQuery. creates a copy of the data in Cloud Storage in TFRecord format, trains the model in Vertex AI Training on that copy, and deploys the model to a Vertex AI endpoint. You have identified the specific version of that model that misclassified, and you need to recover the data this model was trained on. How should you find that copy of the data?

  1. Use Vertex AI Feature Store. Modify the pipeline to use the feature store, and ensure that all training data is stored in it. Search the feature store for the data used for the training.
  2. Use the lineage feature of Vertex AI Metadata to find the model artifact. Determine the version of the model and identify the step that creates the data copy and search in the metadata for its location. Source Reference Answer
  3. Use the logging features in the Vertex AI endpoint to determine the timestamp of the model’s deployment. Find the pipeline run at that timestamp. Identify the step that creates the data copy, and search in the logs for its location.
  4. Find the job ID in Vertex AI Training corresponding to the training for the model. Search in the logs of that job for the data used for the training.

Community Votes

B
100%

100% of anonymous learners picked answer B. Votes are pick records left by other test-takers — they are not the verified answer.

Community Insight

This tests the use of Vertex AI Metadata for artifact lineage tracking, avoiding the trap of relying on manual log searching or pipeline modifications.

To recover the specific training data for a model version, use Vertex AI Metadata lineage. The community confirms that this feature tracks relationships between pipeline artifacts, allowing users to trace a model back to its input data.

Selecting C or D is common because logs are a familiar debugging tool, but they lack the structured relationship tracking needed to reliably identify specific data artifacts compared to Metadata.

Community Discussion (4 comments)

fitri001 👍 4 Selected: B
Vertex AI Metadata Lineage: This feature tracks the relationships between pipeline components and the artifacts they produce. By identifying the model version's lineage, you can pinpoint the specific pipeline run that generated it. Data Copy Step: Within the pipeline run, locate the step responsible for creating the data copy in TFRecord format for training. Metadata Search: Vertex AI Metadata likely stores information about the data copy's location in Cloud Storage, allowing you to access it.
pinimichele01 👍 1 Selected: B
agree with shadz10 and pikachu007
shadz10 👍 2 Selected: B
https://cloud.google.com/vertex-ai/docs/ml-metadata/introduction
pikachu007 👍 3 Selected: B
A. Feature Store: While useful for managing features, it might not store complete training datasets, and modifying the pipeline would not help recover historical data. C. Endpoint Logs and Pipeline Run: This approach involves more manual searching and might be less precise for identifying the exact data copy. D. Training Job Logs: Training job logs might not reliably contain complete data paths or might be purged after a certain period.

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

Submitted for moderation before publishing. Keep it helpful and respectful.

Expert Analysis

Why the Answer Is Correct

Vertex AI Metadata is designed to track the lineage of ML artifacts throughout the MLOps lifecycle. By querying the metadata for the specific model version, you can identify the execution event that created it and trace back to the input artifacts—specifically, the step that created the TFRecord copy. This provides a direct, programmatic link to the exact data location used for training.

Why the Other Options Are Wrong

Option A is incorrect because modifying the pipeline to use Feature Store does not help recover data from a past pipeline run; Feature Store is for serving features, not necessarily storing historical raw training datasets. Option C is inefficient and error-prone, as correlating timestamps between endpoint logs and pipeline runs is manual and less precise than querying metadata. Option D is unreliable because training job logs typically contain metrics and errors, but not necessarily the persistent location of input data artifacts in a structured way.

Community Comment Notes

The community overwhelmingly supports Option B, noting that Vertex AI Metadata lineage is the specific feature for tracking relationships between pipeline components. Comments highlight that while logs (Option C and D) might contain the information, Metadata provides the intended mechanism for artifact discovery. Users also pointed out that Option A is irrelevant to recovering historical data.

Official Reference

https://cloud.google.com/vertex-ai/docs/ml-metadata/introduction

Exam Strategy

When questions ask to trace, recover, or audit the origin of a model or data, immediately look for Metadata or Lineage options. Avoid manual log searches or infrastructure changes for these specific MLOps scenarios.

Related Analysis

You work for a company that builds bridges for cities around

S-Grade

You are training a custom language model for your company us

A-Grade

You trained a model packaged it with a custom Docker contain

S-Grade

Practice All PMLE Questions

Access 65 questions with complete answers and detailed explanations.

View Full PMLE Practice Test →

← Back to PMLE Study Guide