How to train products information in Chat GPT?

Prompting
1

I have a store named ABC that offers a variety of products, primarily related to food and edible groceries. Currently, I am developing a bot that will list products based on dietary preferences. I am in the process of fine-tuning models and feeding SKUs, product names, and their ingredients into the system, utilizing APIs as mentioned in the documentation: https://platform.openai.com/docs/guides/fine-tuning/preparing-your-dataset.

For instance:

User: What is Product123? Assistant: Product123 is a product listed on store ABC, and its ingredients are aa, bb, cc.

Despite my efforts to fine-tune the models using the mentioned APIs, I am encountering issues where the responses generated are either generic or incorrect.

For example, with the prompt:

List all products in ABC store that are Gluten-Free and diabetic-friendly.

I am struggling to obtain accurate responses based on the data fed to my fine-tuned model.

How can I train my model effectively or get started with improving its performance?

2

You should be looking at embeddings for your use case, not fine-tuning. Fine-tuning is predominantly useful for directing models to behave and/or execute tasks in a certain way.

For Q&A type of tasks, embeddings are the way to go.

https://platform.openai.com/docs/guides/embeddings

3