Best finetuning structure in my case

Hello friends,

I’m new to AI, and if my question is wrong, please correct me. I have a page builder, and I want users to be able to change the styles of the page with AI.

What strategy should I have for creating datasets? Should I put all the information about the page and various questions and answers in one file? Or should I divide them into small files? When I have multiple files, how do the data in these files relate to each other? For example, should I put the list of background images in one dataset and create another file for the background colors?

In general, I would appreciate some guidance. Thank you.

Hi Farzam - Welcome to the forum. Could you try to break down your problem a little bit further? What specific aspects are you trying to finetune for?

Thank you for your assistance.

For example, I created a dataset to suggest the page background to the user, and it works well. Now, I want to add the list of images I obtained from Unsplash to suggest background images to the user.

Should I add this list to the previous dataset, or should I create another dataset for these images?

Ok, sorry, not much of help on this specific point. I don’t think it is actually possible to finetune an existing model with picture input but perhaps someone else has looked into that.

In either case, you would normally have to create a separate dataset or create a new model altogether that combines the two aspect under one model.

Thank you. So, in general, if we need multiple separate datasets, should I put them all in one file, right? The possibility of feeding multiple separate files to one model doesn’t exist, and we either need to create different models or put all the data in one dataset. Is that correct?