AI-102 Question 486
Select 3You are tasked with fine-tuning an Azure OpenAI model using your organization's proprietary customer support data. You need to enable the model to respond to customer queries more effectively by incorporating this data. Which steps should you take to achieve this?
- A
Prepare your customer support data in JSONL format and upload it to an Azure Blob Storage container.
- B
Use the Azure OpenAI Studio to create a new model and directly upload the raw CSV data for training.
- C
Set up an Azure Cognitive Search index to enable the model to access relevant documents during runtime.
- D
Use the Azure OpenAI fine-tuning API to train the model with your uploaded JSONL data.
- E
Deploy the fine-tuned model to an Azure App Service to make it accessible via an API.
Show answer and explanation
Correct answers: A, C, D
Explanation
Fine-tuning an Azure OpenAI model with your own data involves several steps: preparing the data in JSONL format, making it accessible (e.g., via Azure Blob Storage), and using the fine-tuning API to train the model. Optionally, integrating Azure Cognitive Search enables the model to retrieve relevant documents during runtime, enhancing its context-awareness. Deploying the model can follow once these steps are completed.
- A. Correct.
Correct. JSONL (JSON Lines) is the required format for training data when fine-tuning an Azure OpenAI model. Uploading the prepared data to an accessible location, like Azure Blob Storage, is a key step.
- B. Incorrect.
Incorrect. While CSV is a common data format, Azure OpenAI fine-tuning requires JSONL format. Raw CSV data cannot be directly used for training.
- C. Correct.
Correct. Setting up an Azure Cognitive Search index allows the model to retrieve relevant documents dynamically during runtime, improving its ability to respond to queries.
- D. Correct.
Correct. The Azure OpenAI fine-tuning API is used to train the model with your data. This step is essential to incorporate your proprietary data into the model's training.
- E. Incorrect.
Incorrect. While deploying the fine-tuned model is a logical next step, Azure App Service is not required for this purpose. The model can be deployed directly via Azure OpenAI or another hosting service.