Multiple Choice Flashcards on 02 Databricks Certified Machine Learning Profes...

Question Type

Reader

Answer workspace

Explanation appears after submit

FREE AI FEEDBACK

Why this matters

Review the keyed rationale after answering so the correct AWS service choice and each distractor stay connected.

On demand Read aloud

0 Recent saved answers

Recent Databricks Certified Machine Learning Professional Answers

Active recall loop

Answer, inspect the rationale, then repeat.

Large answer targets and a focused review step keep the current AWS objective at the center of the session.

Answer Review Repeat

Flashcard Practice Databricks Certified Machine Learning Professional Multiple Choice

A team is designing training for a large language model that cannot fit on one GPU, and the corpus is large enough to require high throughput. The cluster has fast links inside each GPU node and slower links between nodes. Which strategy is most defensible?

Choose the Best Answer

Use pure data parallelism and rely on the corpus size to occupy every GPU. Use pure model parallelism across the cluster with one active replica. Use hybrid parallelism aligned to model size and topology. Train one checkpoint per node and ensemble the final models.

Load another random question

More

Question

Choose the Best Answer

How can I help?