Software Engineering Courses (SECourses)•5mo ago

How many epochs would you recommend for training 50 images for a 1024x1024 close-up face?

KawixOP•8/21/25, 5:34 PM

Small question: if I train my fine-tuned flux model, could I use it for inpainting, or do I need other parameters in koyas_ss to make it work a bit like flux fill?

KawixOP•8/21/25, 5:39 PM

Should I do my training with flux fill instead of flux dev?

KKawix Small question: if I train my fine-tuned flux model, could I use it for inpainti...

skyrrr•8/21/25, 5:57 PM

You can use your finetuned checkpoint to do impainting
Finetuning of flux fill isn't possible

KKawix How many epochs would you recommend for training 50 images for a 1024x1024 close...

skyrrr•8/21/25, 5:58 PM

Between 100 and 200. Around 150 is often the sweet spot with flux.dev, it depends on your dataset. I would recommend to save every 20 or 25 epochs, and compare on swarmui using the grid generation tool

Sskyrrr Between 100 and 200. Around 150 is often the sweet spot with flux.dev, it depend...

KawixOP•8/21/25, 6:20 PM

ok thx i will do this and follow the tutorial

KawixOP•8/21/25, 6:23 PM

@skyrrr By any chance, are you familiar with face detection and Comfy UI? I'm struggling to create a workflow that would detect faces with Yolofacev8 or RetinaFace and then inpaint the face area, which would allow me to regenerate that part with my trained fluxfinetuned. But I can't find a good node. I tried installing “ComfyUI Yolov8” but I got an import error, so I guess it's not compatible. There's no information about this on GitHub.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/21/25, 6:28 PM

i'm scared to use this part of kohya, dont wanna waste a night of training and the model files werent saved anywhere...

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/21/25, 6:28 PM

but would be practical for sure

KawixOP•8/21/25, 6:32 PM

i scared to haha

KawixOP•8/21/25, 6:33 PM

i didi'nt do the tutorial yet

KawixOP•8/21/25, 6:33 PM

but i wish everything will work for my flux fine tuned

KawixOP•8/21/25, 6:33 PM

because time is money on cloudGPU :((

KawixOP•8/21/25, 6:49 PM

It's hell trying to find a simple node that allows me to simply only mask a face.

KawixOP•8/21/25, 7:25 PM

something like that

KawixOP•8/21/25, 7:25 PM

but more simple i wish

KKawix How many epochs would you recommend for training 50 images for a 1024x1024 close...

Furkan Gözükara SECourses•8/21/25, 7:29 PM

min 200

Furkan Gözükara SECourses•8/21/25, 7:29 PM

and compare

KKawix Small question: if I train my fine-tuned flux model, could I use it for inpainti...

Furkan Gözükara SECourses•8/21/25, 7:29 PM

it works right away as flux fill

Furkan Gözükara SECourses•8/21/25, 7:29 PM

you can do inpainting

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭i'm scared to use this part of kohya, dont wanna waste a night of training and t...

Furkan Gözükara SECourses•8/21/25, 7:29 PM

i tested once or twice never worked

KawixOP•8/21/25, 7:49 PM

Thx Furkan ! Do u kwn something good for me to face detection and mask ?

KawixOP•8/21/25, 7:50 PM

I thought I had found the perfect thing for me, but apparently comfyui has decided to give me more errors, as always.

KawixOP•8/21/25, 7:59 PM

ok i think it's maybe beacause python 3.10 only not above

Sskyrrr Between 100 and 200. Around 150 is often the sweet spot with flux.dev, it depend...

JonkoXL•8/21/25, 9:03 PM

How many repeats and learning rate for 150 images at 20-25 epochs?

KKawix ok i think it's maybe beacause python 3.10 only not above 😡

Furkan Gözükara SECourses•8/21/25, 9:19 PM

that is why i still use python 3.10.11

Furkan Gözükara SECourses•8/21/25, 9:19 PM

most widely supported version

JJonkoXL How many repeats and learning rate for 150 images at 20-25 epochs?

Furkan Gözükara SECourses•8/21/25, 9:19 PM

use our presets dont change

JonkoXL•8/21/25, 9:20 PM

I'm a tinkerer too

JonkoXL•8/21/25, 9:20 PM

I guess I'll have to grep search then if I want the answer

JJonkoXL How many repeats and learning rate for 150 images at 20-25 epochs?

skyrrr•8/21/25, 9:37 PM

You should follow @Dr. Furkan Gözükara method, it does wonders. Set repeats to 1. Train more epochs, at least 150

KKawix Thx Furkan ! Do u kwn something good for me to face detection and mask ?

Furkan Gözükara SECourses•8/21/25, 10:01 PM

our app can crop faces

Furkan Gözükara SECourses•8/21/25, 10:01 PM

https://www.patreon.com/posts/120352012

Patreon

Ultimate Image Processing APP : Batch Cropping, Zooming In, Resizin...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

Furkan Gözükara SECourses•8/21/25, 10:01 PM

and SAM can mask them

Furkan Gözükara SECourses•8/21/25, 10:01 PM

we have SAM too

FFurkan Gözükara SECourses https://www.patreon.com/posts/120352012

KawixOP•8/21/25, 10:56 PM

ok thx i can't use this in comfyui ?

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/21/25, 11:21 PM

what do you set guidance scale to when fine tuning a flux de destilled model?

筋

筋斗云•8/22/25, 6:15 AM

I actually quite recommend an online ComfyUI running platform called Runinghub. You don't need to worry about ComfyUI environment errors. Basically, the latest technologies can be experienced or there are workflows available on it in a timely manner. The vast majority of them can be used directly and can also be downloaded. Although this may seem like an advertisement, I've actually subscribed to an annual membership myself. I use a 48G 4090 to run workflows and applications. If anyone needs it and it's appropriate, I can send the link.

筋

筋斗云•8/22/25, 6:20 AM

KKawix ok thx i can't use this in comfyui ?

Furkan Gözükara SECourses•8/22/25, 8:38 AM

you can use

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭what do you set guidance scale to when fine tuning a flux de destilled model?

Furkan Gözükara SECourses•8/22/25, 8:38 AM

3.5

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 9:42 AM

I just used default config. Will see how that goes. The 48gb one

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 11:08 AM

has anyone completed a wan 2.2 lora or fine tuning yet on a person? how do the results look?

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭has anyone completed a wan 2.2 lora or fine tuning yet on a person? how do the r...

cyberbol•8/22/25, 11:17 AM

Mostly problems with eyes...

I looking, thinking about right way how to train lora for wan2.2, specially what dataset to use. My doughter ask me to make for her LORA.

I dont know what aproach to do. Should I for HIGH noise use smaller resolution video clips, and for LOW NOISE use high quality photos.

Or maybe use same mixed dataset, videos and fotos, for both HIGH and LOW.

Hard for me to find any info wich way works best.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 6:47 PM

Epoch 80, flux krea fine tune. 25 pics

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 6:48 PM

Epoch 120 turned out to be best

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 6:48 PM

Used @Dr. Furkan Gözükara fine tune 48gb config

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•8/22/25, 7:13 PM

Wut? My photo was 100 % sfw?