フリー問題

問題文

You want a deployed multimodal model to describe what is wrong in a photograph of a damaged shipment. How do you supply the image?

選択肢

1. Upload the image to storage and put only the file path in the prompt text
2. Train a custom model on your shipment photographs first, because a deployed model will reject any image it has not already seen during training
3. Include the image as part of the prompt content alongside the text instruction in the same request
4. Convert the image to a text description first, then send only the description