Fine-Tuning — Google Cloud ML Engineer Practice Questions

Fine-tuning is the process of continuing to train a pre-trained model on a domain-specific dataset so its weights adapt to a new task or distribution. The ML Engineer exam distinguishes fine-tuning from transfer learning in that fine-tuning typically updates more of the model's weights rather than training only a new output head. On Google Cloud, fine-tuning is supported for foundation models through Vertex AI, including supervised fine-tuning for generative models in Vertex AI Studio. Exam questions often require candidates to choose between full fine-tuning, parameter-efficient fine-tuning methods, and prompt engineering based on dataset size, compute budget, and performance targets.

Free questions on fine-tuning

More fine-tuning questions in the full bank

• When fine-tuning a pre-trained language model on Vertex AI for domain-specific tasks, what is the risk of aggressive learning rates? Unlock answer & explanation →
• What is transfer learning and when should you use it? Unlock answer & explanation →
• You are fine-tuning a pre-trained EfficientNet model on a small custom image dataset using Vertex AI. The model was pre-trained on ImageNet. Which layer-freezing strategy is most appropriate to avoid overfitting while leveraging the learned features? Unlock answer & explanation →