CLOUD-DIGITAL-LEADER · Question #393
An organization has hired a team of data scientists and developers. They want to create unique value in their business by coding an advanced machine learning model in Vertex AI Workbench. Which servic
The correct answer is A. Custom training. Vertex AI Custom Training is the service that allows data scientists to write and run their own ML code for training models with full control over the training logic.
Question
An organization has hired a team of data scientists and developers. They want to create unique value in their business by coding an advanced machine learning model in Vertex AI Workbench. Which service should the organization use to train the model?
Options
- ACustom training
- BAutoML
- CCompute Engine
- DPrebuilt APIs
How the community answered
(21 responses)- A81% (17)
- B5% (1)
- C10% (2)
- D5% (1)
Why each option
Vertex AI Custom Training is the service that allows data scientists to write and run their own ML code for training models with full control over the training logic.
Vertex AI Custom Training lets teams package their own training code - written in TensorFlow, PyTorch, scikit-learn, or other frameworks - and run it at scale using Google Cloud managed infrastructure. Since the team is coding an advanced model in Vertex AI Workbench, Custom Training is the natural companion service for executing that code at scale with configurable compute, including TPUs and GPUs.
AutoML automates model building without requiring custom code, which is the opposite of what the organization needs since they are writing their own advanced model.
Compute Engine is a general-purpose VM service not purpose-built for ML training workflows and lacks the MLOps integrations of Vertex AI.
Prebuilt APIs provide access to pre-trained Google models for common tasks and do not support training custom models.
Concept tested: Vertex AI Custom Training for custom ML model development
Source: https://cloud.google.com/vertex-ai/docs/training/overview
Topics
Community Discussion
No community discussion yet for this question.