Exam GES-C01 Topic 1 Question 33 Discussion

To use a fine-tuned model for inference, you can call the 'COMPLETE' (or 'AI_COMPLETE') LLM function with the name of your fine-tuned model. For structured output, you can specify a JSON schema using the 'response_format argument with 'AI_COMPLETE. For OpenAI (GPT) models, the 'additionalProperties' field must be set to and the 'required' field must contain the names of every property in the schema. For the most consistent results, setting the 'temperature' option to 0 when calling 'COMPLETE (or 'AI_COMPLETE) is recommended. JSON schema guidelines also state that object definitions should be placed at the top level of the schema, specifically under the 'definitions' or '$defs key, and Snowflake's validation strictly enforces this structure. Cortex LLM functions do not support dynamic tables. While fine-tuned models appear in the Snowsight UI of the Model Registry, they are not managed by the Model Registry API. Usage of Cortex functions can be tracked using views like "CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY', but this is for tracking costs/usage, not for real-time model monitoring in the context of Model Registry.