About training data

#50
by shirleyjie - opened

I would like to ask you about the training data. Did you download the training data from the Internet? I would like to ask where the data source comes from, and how much training data is it?

This is an irrelevant question, as you'd need to ask this of Stable DIffusion AI themselves, not just Lykon.

This is an irrelevant question, as you'd need to ask this of Stable DIffusion AI themselves, not just Lykon.

Hello, thank you for your answer. What I actually want to ask is the amount of data needed for fine-tuning based on the basemodel, and how much data is used to get a model as high-quality as dreamshaper. I will be very grateful for all the answers

OH sorry LOL.
My bad i wouldn't know what Lykon used, i know personally what i've used--
Anime models a lot of the time vary - i know Linaqruf used about 25k images or something over 25k steps--
I used at the start for Osea XL only about 200images - but it also was pretty weak because i'm used to stronger models.

It is partially sort of preference, Lykon will probably have opinions I dont - i apologize i just watch their stuff like a hawk because i like their content :P

OH sorry LOL.
My bad i wouldn't know what Lykon used, i know personally what i've used--
Anime models a lot of the time vary - i know Linaqruf used about 25k images or something over 25k steps--
I used at the start for Osea XL only about 200images - but it also was pretty weak because i'm used to stronger models.

It is partially sort of preference, Lykon will probably have opinions I dont - i apologize i just watch their stuff like a hawk because i like their content :P

Oh, thank you very much for your answer. I would like to ask if the data on your side is collected by yourself, or is it obtained from model production?

Lykon changed discussion status to closed

Sign up or log in to comment