Is there any way to get automatic questions and answers for my dataset. For example I have a review rating dataset for multiple cities. Expected output is when I send the dataset it should be able to generate all possible prompts and completions on my data and return me prompt completion pairs. I will use the prompt completion pairs for fine-tuning a text davinci model.
Thanks in advance
Is there a specific reason you want to use fine-tuning instead of embedding your dataset?
The latter will be lower cost and allow you to use the latest models, like that ChatGPT API and text-davinci-003
.
If you’re interested, I made a couple diagrams showing how this would work with embeddings: Q&A with ChatGPT & Embeddings - Basic & HyDE Examples - WFH Brian
For my use case finetuning will be the best approach . But still I am figuring a way to create prompts and completions from my tabular data