1Z0-184-25 Question 105
Select 2Your organization needs to generate vector embeddings for 50 million product descriptions stored in Oracle Cloud Infrastructure Object Storage. These embeddings will be used in your Oracle Database for advanced similarity searches. Due to performance and compute constraints, you have decided to generate these embeddings outside of the database. Which two approaches provide a suitable and scalable solution for generating and storing these vector embeddings while meeting your performance requirements?
- A
Use Oracle Data Science to run a custom Python script on GPU-enabled shapes, reading data from Object Storage, generating embeddings, and writing them back to Object Storage or directly into the database.
- B
Leverage Oracle Functions with minimal local storage and memory, invoking the function repeatedly to process large volumes of data in real time.
- C
Use Oracle AI Services � Language to perform batch embeddings on text data from Object Storage, then store the resulting embeddings in the Oracle Database for analysis.
- D
Rely solely on PL/SQL in the Oracle Database to execute Java or Python stored procedures for generating embeddings within the database environment.
Show answer and explanation
Correct answers: A, C
Explanation
For high-volume text embedding generation, best practices often recommend using specialized services or scalable data science environments outside the database. Oracle Data Science can provide GPU-enabled compute for large-scale embedding generation, while Oracle AI Services � Language offers batch text-embedding capabilities that can be invoked using data from OCI Object Storage. Both approaches allow you to offload intensive computation from the Oracle Database, then store embeddings for future use (e.g., similarity searches in Oracle Database). Refer to the Oracle Data Science documentation and Oracle AI Services � Language documentation for detailed instructions on setting up batch text analytics workloads outside the database.
- A. Correct.
Correct. Oracle Data Science is well-suited for large-scale machine learning tasks. Running a custom Python script (potentially on GPU shapes) lets you efficiently handle large data stored in Object Storage, generate embeddings, and store those embeddings externally or load them back into the Oracle Database once completed.
- B. Incorrect.
Incorrect. While Oracle Functions can handle smaller tasks and stateless operations effectively, it is not typically recommended for large-scale batch embeddings (e.g., 50 million records). Functions have limited memory and execution time, making them less optimal for large, long-running operations.
- C. Correct.
Correct. Oracle AI Services � Language can generate text embeddings in batch, making it feasible to process larger datasets outside the database. You can then store the embeddings in the Oracle Database for further similarity searches or analytics.
- D. Incorrect.
Incorrect. This explicitly generates embeddings inside the database, which contradicts the stated requirement to do it outside the database for performance reasons. Additionally, relying solely on in-database procedures can introduce higher overhead for large-scale data processing.