Software Engineering Courses (SECourses)•8mo ago

Anyone else suddenly having issues fine tuning with kohya on Massed compute? I've done this like 50

Anyone else suddenly having issues fine tuning with kohya on Massed compute? I've done this like 50 times now the past 3 months and am suddenly getting errors

JamiewheltonOP•5/4/25, 4:39 AM

anyone have any idea what I might be doing wrong?

Screenshot_2025-05-04_at_12.39.19_PM.png

Screenshot_2025-05-04_at_12.39.47_PM.png

JamiewheltonOP•5/4/25, 6:36 AM

Looks like somethings up with the new repo updates to the models

FFurkan Gözükara SECourses yes

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/4/25, 2:20 PM

Thank you Dr. And grabbing the 1024px ones will be sufficient. No point going higher. And you already cropped to make the person occupy most of the photos yes? So I don't have to run your ultimate image process on them.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/4/25, 5:43 PM

doing my first sdxl fine tune with using regimages. i had 54 training images, so it looks like it's using 55 reg images per epoch, does this look right? don't wanna waste 25h

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/4/25, 5:43 PM

@Furkan Gözükara SECourses

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/4/25, 9:17 PM

Time changed to 6h. Nice

JJamiewhelton Anyone else suddenly having issues fine tuning with kohya on Massed compute? I'v...

Furkan Gözükara SECourses•5/4/25, 10:59 PM

just fixed today

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭doing my first sdxl fine tune with using regimages. i had 54 training images, so...

Furkan Gözükara SECourses•5/4/25, 10:59 PM

looks good

Furkan Gözükara SECourses•5/4/25, 10:59 PM

54 + 55 = 99

Furkan Gözükara SECourses•5/4/25, 11:00 PM

you do only 10 extra :d

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/5/25, 6:06 AM

Uh Dr. 54 + 55 = 109. Thanks

maybe I'll try masked training too.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/5/25, 6:15 AM

Interesting, this person fine tunes a new woman by mixing 3 real women like described here. Is this possible with merging or?

PabloMoises•5/5/25, 9:46 AM

Hi!

I need help with a couple of things, let’s see if someone can assist me:

For an average of 200 images in a fine-tuning using Flux, how many epochs would you recommend?

To train on a multi-GPU VM, is there anything special that needs to be done? Do I need to modify anything or is nothing required?

Thank you very much!

PPabloMoises Hi! I need help with a couple of things, let’s see if someone can assist me: F...

Furkan Gözükara SECourses•5/5/25, 2:24 PM

up to 100 i recommend. but probably you will get best around 70 80

PPabloMoises Hi! I need help with a couple of things, let’s see if someone can assist me: F...

Furkan Gözükara SECourses•5/5/25, 2:24 PM

multi gpu works on 80 gb gpus for dreambooth or 48 gb for lora

Furkan Gözükara SECourses•5/5/25, 2:24 PM

we have config and tutorial for it

Furkan Gözükara SECourses•5/5/25, 2:25 PM

https://youtu.be/-uhL2nW7Ddw

YouTubeSECourses

Blazing Fast & Ultra Cheap FLUX LoRA Training on Massed Compute & R...

Unlock the power of FLUX LoRA training, even if you're short on GPUs or looking to boost speed and scale! This comprehensive guide takes you from novice to expert, showing you how to use Kohya GUI for creating top-notch FLUX LoRAs in the cloud. We'll cover everything: maximizing quality, optimizing speed, and finding the best deals. With our exc...

Furkan Gözükara SECourses•5/5/25, 2:25 PM

also i think how multi gpu linked matters. i got error on pci express multi gpu

FFurkan Gözükara SECourses up to 100 i recommend. but probably you will get best around 70 80

PabloMoises•5/5/25, 2:36 PM

One of the full finetuning training I tried, I used 300 epochs with exactly 200 images. Maybe that’s too much?

PPabloMoises One of the full finetuning training I tried, I used 300 epochs with exactly 200 ...

Furkan Gözükara SECourses•5/5/25, 3:17 PM

probably unnecessary

Furkan Gözükara SECourses•5/5/25, 3:17 PM

but sometimes i go high and compare checkpoints

Furkan Gözükara SECourses•5/5/25, 3:23 PM

just fixed google colab of this tutorial

Furkan Gözükara SECourses•5/5/25, 3:23 PM

https://youtu.be/mnCY8uM7E50

YouTubeSECourses

Transform Your Selfie into a Stunning AI Avatar with Stable Diffusi...

Our Discord : https://discord.gg/HbqgGaZVmr. How to do free Stable Diffusion DreamBooth training on Google Colab for free. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on

https://www.patreon.com/SECourses

Playlist of Stable Diffusion Tutorials, Automatic1111 and ...

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/5/25, 6:09 PM

@Furkan Gözükara SECourses what is the result of training 15 pics of woman A then 10 pics of woman B, ohwx woman as token for every pic? Result would be like a mean of their looks?

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/5/25, 6:10 PM

Training images consisting of two women, but all pics are solo. 15 pics of A, 10 pics of B

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭<@205854764540362752> what is the result of training 15 pics of woman A then 10 ...

Furkan Gözükara SECourses•5/5/25, 10:33 PM

mixed

Furkan Gözükara SECourses•5/5/25, 10:33 PM

bled

Llerois1 I am pleased to inform you that the dreambooth training has produced excellent r...

JAV_Pomoe•5/6/25, 3:36 AM

I got the same result. If this is a specification of Flux, it is a big problem. I hope there is some way to avoid it by ingenuity during training. I will try to get the same result by using a caption file.

FFurkan Gözükara SECourses mixed

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/6/25, 8:07 AM

i dont' fully understand, will the result be a new consistent look?

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•5/6/25, 8:07 AM

Dr. We are hopeful that you will be the first to teach us how to fine tune on Hi Dream Dev, I hear Dev actually is better than the full version. @Furkan Gözükara SECourses

Original message was deleted

JAV_Pomoe•5/6/25, 8:32 AM

This is the first time I've heard of this model. What do you like about it? Where can I get more information about Chroma?

Original message was deleted

JAV_Pomoe•5/6/25, 8:36 AM

Oh, thank you very much!

Sakkimuch•5/6/25, 10:37 AM

Im new on LORA training and i have a question: what is if i want to train a full body (not only face and hairs and little bite of the upper body) LORA of a person. Can i train it on as example 1350 x 1080 or 1920 x 1080 or i need to train it on 1024 x 1024 too and crop the images? And can i generate other resolutions after training? Sorry for this dumb questions but i never trained a lora and in the tutorial all is trained and cropped by 1024 x 1024. Thanks alot

Sakkimuch•5/6/25, 11:45 AM

Ok, thank you. But the image sizes have no direct influence on the images I can generate later, right? Because the model only learns the object with the images, for example, and not the size. So if, for example, I have 50 images with a resolution of 1080 x 1350 that contain everything that is important for the LORA, I can also generate other resolutions later, such as 896 x 1152 (portrait), right? Because LORA training only learns the object?

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭Dr. We are hopeful that you will be the first to teach us how to fine tune on Hi...

Furkan Gözükara SECourses•5/6/25, 3:09 PM

i plan to fully research

SSakkimuch Im new on LORA training and i have a question: what is if i want to train a full...

Furkan Gözükara SECourses•5/6/25, 3:10 PM

you can train different aspect ratios but best is image resolution being 1 megapixel

Furkan Gözükara SECourses•5/6/25, 3:10 PM

also i train 1024x1024 and can generate any aspect ratio

Furkan Gözükara SECourses•5/6/25, 3:10 PM

but training with your target aspect ratio helps

SSakkimuch Ok, thank you. But the image sizes have no direct influence on the images I can ...

Furkan Gözükara SECourses•5/6/25, 3:10 PM

make sure everything is at least 1 mega pixel

Furkan Gözükara SECourses•5/6/25, 3:11 PM

also try to make all same aspect ratio

Sakkimuch•5/6/25, 3:16 PM

Ok thx all my images (i have 63 with different clothes, impressions etc.) are 1080 x 1350 if i train this, the model are not cropping them automatically to 1024 by 1024 and some details gone right? sry im really new at this thematic?

SSakkimuch Ok thx all my images (i have 63 with different clothes, impressions etc.) are 10...

Furkan Gözükara SECourses•5/6/25, 3:18 PM

that sounds weird ratio but you can train with it

Furkan Gözükara SECourses•5/6/25, 3:18 PM

no issues

Furkan Gözükara SECourses•5/6/25, 3:18 PM

for more details i recommend after training do 2x upscale

Furkan Gözükara SECourses•5/6/25, 3:18 PM

this video shows upscale : https://youtu.be/fTzlQ0tjxj0

YouTubeSECourses

Master Local AI Art & Video Generation with SwarmUI (ComfyUI Backen...

If you want to generate the very best AI videos and images on your Windows computer locally this is the tutorial that you were looking for. Literally 1-click to install most powerful and advanced generative AI interface SwarmUI (with Flash Attention, Sage Attention, Triton, DeepSpeed, xFormers, RTX 5000 series perfect compatibility) and download...

FFurkan Gözükara SECourses that sounds weird ratio but you can train with it

Sakkimuch•5/6/25, 3:21 PM

what ratio you recommend for full or half body training or is it best to crop them manually to 1024 x 1024 with the most of body and face what is possible

Sakkimuch•5/6/25, 3:22 PM

if it gets best quality with 1024px i cropped it no problem

SSakkimuch what ratio you recommend for full or half body training or is it best to crop th...

Furkan Gözükara SECourses•5/6/25, 3:24 PM

i trained 1024x1024 it works perfect

Furkan Gözükara SECourses•5/6/25, 3:24 PM

as long as your dataset good it will work

Sakkimuch•5/6/25, 3:28 PM

ok last question: would you recommend captions. I think I read in one of your posts that it doesn't help with LORAs, is that still the case?

Anyone else suddenly having issues fine tuning with kohya on Massed compute? I've done this like 50

Similar Threads

Similar Threads

Similar Threads