I need that the fine tune model answer over my training data, its posible? or ¿ Do I need use embeddings-based search?

API
, ,
1

I need to relate the email requests with the data in my database because they don’t have the same thing, the idea I have is to get the fine-tuning model to have the data from my database and that when I send an email the model managed to relate information and return the objects I need or values so that I can generate a correct order in the database.

I have almost everything ready but the part about using Openai is not clear, I don’t know if the trained model will remember the data used in the training or I have to use another method, I also expected to continue training the model with future errors to get the model to learn from a mistake by not relating information and that in future emails it will detect the problem and use previous relationships.

thanks for make an answer.

2

It really depends on the outcome. From your question, its not clear on what you want to achieve - can you give an example please?

3

For example, I have data from my database and I need the fine-tuning to answer questions about my own data, but the fine-tuning may not remember the data used in the training file, my idea is that Openai creates a relationship between an email and data in my database, emails are product orders and I want it to be a more automatic process.
So if I want to use fine-tuning how the training file is structured or I need to use embeddings.
The result will be a JSON using OpenAI’s function call.

4

Correct. For your use case, its better to use Embeddings. Fine tuning is better if you want to get the formatting, tone of voice etc accurate.