Finding all my models or profile elsewhere than Civitai is without my consent !
Please don't repost my files. Send people to my Civitai profile, thanks :)
-
ROOMS is a new group of loras more focused on indoor spaces.
--
This is a background Lora with cyberpunk vibes. This is the Z Image Turbo edition.
Due to 100% failed Z Image Turbo lora training using Civitai services (I have 17 failed and some still running for days), I had to learn really quickly how to use Ai Toolkit... so...
Yay ! This is my first offline trained Z image Turbo lora !
Huh, it took about 6 hours to train...
FOR FUN AND TESTING :)
--
Cyberpunk Bedrooms 01 and more :
Bedrooms, living rooms, other spaces and with Z Image great prompt adherence, you can do whatever you want ;)
Two kinds of prompting in the dataset :
natural langage
tags
--
HOW TO USE :
Strength : 1 tested
Steps : 9 to 20 tested
Checkpoints : Any but I used the Base Fp8 and Base BF16 models.
Adetailer : your choice
Sampler : Euler (at 9-20 steps) or DPM++ 3M SDE (at 20 steps)
Scheduler : NORMAL
--
Trigger word :
CYBR1
--
See pics for prompts ;)
--
No effects. Due to training times and memory limitations, I had to use the smallest dataset.
--
Have fun :)
Description
Made by Z1os on civitai. https://civitai.com/user/Z1os
Finding all my models or profile elsewhere than Civitai is without my consent !
If you see this model or my profile elsewhere, then this site is a stealer.
--
Please don't repost my files. Send people to my Civitai profile, thanks :)
FAQ
Comments (5)
6 hours of training with which video card?
I have a RTX 3070 with 8 gb of vram and DDR4 pc. But I borrowed a RTX 3060 so I could have 12Gb of vram to make the training.
I have now found a new way of reducing training times, by just changing parameters. Takes about 2 hours now with the same card.
@Z1os I'm curious to see how much I would be able to do with my 12GB 4070ti. Is it just a matter of VRAM or does the driver, architecture, bus, CUDA and tensor cores, etc. also count? Is there any way you can provide a guide on these new parameters to use to improve calculation times? Does it make much of a difference in terms of training quality?
@NoMarzipan8994381 I'm still testing things. I haven't found a stable "profile" I can use every time. The important thing in training is the card and Vram. You have a 4070Ti... I guess you could train nice loras in just an hour. I have to do some more tests but ok, I'll do a guide with what I use.