AWS Certified AI Practitioner Practice Exam

A company uses Amazon Bedrock for an AI search tool and wants to fine-tune the model. Which strategy is appropriate?

1. Provide labeled data with the prompt field and the completion field

2. Prepare the dataset using a .txt file in .csv format

3. Purchase Provisioned Throughput for Amazon Bedrock

4. Train the model on journals and textbooks

The correct answer is: Provide labeled data with the prompt field and the completion field

Providing labeled data with both the prompt field and the completion field is the appropriate strategy for fine-tuning a model using Amazon Bedrock. Fine-tuning involves adjusting a pre-trained model on a new dataset, and labeled data is crucial for this process. In this context, the prompt field consists of the inputs to the model, while the completion field represents the expected outputs or responses. This clear pairing allows the model to learn the relationship between specific prompts and their desired outputs effectively. By using this structured data, the model can adjust its parameters to improve its accuracy and relevance for the specific tasks at hand, ultimately enhancing the performance of the AI search tool. Contextually, the other suggestions do not align with best practices for fine-tuning. Preparing a dataset in a .txt file in .csv format might not provide the necessary structured approach that labeled data offers for training. Purchasing Provisioned Throughput addresses performance at scale, which may be beneficial but is not directly related to the fine-tuning process itself. Training the model on journals and textbooks could contribute to general knowledge but does not focus on the specific, targeted adjustments needed for the AI search tool's intended function.