Databricks Generative AI Engineer Associate Question 176
Select 3You are tasked with deploying a generative AI application on Databricks that generates product descriptions based on input data. The application uses a fine-tuned large language model (LLM) and needs to handle high request volume with minimal latency. Which steps should you follow to ensure optimal deployment in this scenario?
- A
Use Databricks Model Serving to host the fine-tuned model and handle inference requests.
- B
Package the application as a REST API using Databricks Jobs and deploy it outside Databricks for scalability.
- C
Enable autoscaling for your Databricks cluster and optimize the cluster size based on anticipated traffic.
- D
Deploy the application using Databricks Workflows to automate model loading and API serving.
- E
Implement request batching and caching mechanisms to handle high request volume efficiently.
Show answer and explanation
Correct answers: A, C, E
Explanation
To deploy a generative AI application with high performance and scalability on Databricks, it's essential to use Databricks Model Serving for low-latency inference, enable cluster autoscaling to handle varying traffic loads, and implement request batching and caching to optimize performance. These steps ensure the application is robust and efficient under high demand.
- A. Correct.
Databricks Model Serving is the recommended approach for hosting models and handling inference requests directly on Databricks, ensuring low-latency responses.
- B. Incorrect.
Deploying the application outside Databricks using Databricks Jobs is not optimal for this scenario as it adds complexity and may lead to increased latency.
- C. Correct.
Enabling autoscaling ensures that the cluster can dynamically adjust resources based on incoming traffic, helping to maintain performance during high request volumes.
- D. Incorrect.
Databricks Workflows is primarily used for orchestrating tasks and pipelines, rather than directly serving models or handling real-time API requests.
- E. Correct.
Request batching and caching are best practices for optimizing performance when dealing with high request volumes, as they reduce redundant processing and improve throughput.