Snowflake SnowPro® Specialty: Gen AI Certification 認定 GES-C01 試験問題:

1. A security-conscious data scientist in an Azure East US 2 (Virginia) account wants to fine-tune a mistral -7b model for a specific text summarization task and then deploy it for real-time inference using the Cortex REST API. The mistral-7b base model is natively available for fine-tuning in Azure East US 2 (Virginia). For subsequent inference using the fine-tuned model, they need to understand the regional and cross-region inference considerations. Which of the following statements are correct?

A) Option E
B) Option A
C) Option C
D) Option D
E) Option B

2. A development team is preparing to deploy a new Retrieval-Augmented Generation (RAG) application written in Python. They intend to use Snowflake AI Observability with the TruLens SDK to capture detailed logs and traces for debugging and performance analysis. Which of the following configurations are essential prerequisites for enabling this logging capability effectively?

A) Option E
B) Option A
C) Option C
D) Option D
E) Option B

3.

A) Option E
B) Option A
C) Option C
D) Option D
E) Option B

4. A Snowflake data engineering team needs to implement a system where they can call the SNOWFLAKE .CORTEX. COMPLETE function using the model, which is hosted in a different region than their primary Snowflake account. They also want to ensure that only llama3.1-8b approved models are callable. Which of the following configurations or privileges are necessary to achieve these requirements?

A) Option E
B) Option A
C) Option C
D) Option D
E) Option B

5. A financial institution is building a Document AI model to process loan applications. They need to extract all applicant names (which can be multiple, forming a list), classify the application as 'Approved' or 'Rejected' based on various internal criteria (not explicitly stated in the document), and normalize diverse date formats (e.g., '1st Jan 2023', '01/01/2023') to 'YYYY-MM-DD'. Which of the following approaches represent 'best practices' for defining data values and training the Document AI model for these complex extractions? (Select ALL that apply.)

A) When defining data values for combinations of values, arrays, or nonstandard formats, fine-tuning the model with annotations that 'show' the expected result is generally more effective than 'telling' it via complex prompt engineering.
B) For applicant names, define a single value 'applicant_name' and rely solely on advanced prompt engineering to instruct the model on how to identify and list all names, without providing specific examples through annotations.
C) To normalize diverse date formats to 'YYYY-MM-DD', train the model with sufficient annotations, showing the desired output format for various input date formats, allowing Document AI to learn the normalization internally.
D) For extracting lists of items from tables, use individual data values for each column and then merge these lists in the pipeline, ensuring adequate training with sample data, including NULL values and correct order, for reconstruction.
E) To classify applications as 'Approved' or 'Rejected', define a data value like 'What is the application status?' and ensure the training dataset includes annotated examples for both 'Approved' and 'Rejected' classes.

質問と回答：