hello sorry to ask but what training parameters should i use for training 56 image model
hello sorry to ask but what training parameters should i use for training 56 image model
Falso if im out of space currently on massed compute how can i add gpu without losing all of my data?
Fsame as our config
Fstill do up to 200 epoch if you have time
Fsadly you cant as far as i know @NicB
S@Dr. Furkan Gözükara If I want to do a full finetune of Flux is this the best tool? https://github.com/bmaltais/kohya_ss/ Do you have a sample configuration? Thanks!
GitHub
Contribute to bmaltais/kohya_ss development by creating an account on GitHub.
Fyes we have
FYouTubeSECourses
If you want to train FLUX with maximum possible quality, this is the tutorial looking for. In this comprehensive tutorial, you will learn how to install Kohya GUI and use it to fully Fine-Tune / DreamBooth FLUX model. After that how to use SwarmUI to compare generated checkpoints / models and find the very best one to generate most amazing image...
Fup to date configs here both lora and dreambooth / fine tuning : https://www.patreon.com/posts/click-to-open-post-used-in-tutorial-112099700
Patreon
Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon
JNeed good data augmentation tips for making a more realistic dataset of an AI person. Specifically more skin detail on the face
Jbest EPOC for 39 images?
Jmy images are between 1536x1536 and 2152 x 2152 res
is that an issue? flux will downscale, but is there a way I can get more out of these higher rep images in my fine-tuned model?
is that an issue? flux will downscale, but is there a way I can get more out of these higher rep images in my fine-tuned model?
Fdo at least 200 and compare
F1024 working best
Fbigger res not helping much
Fyou can agument your data and add zoomed in face images etc
SHey, can we try a full finetune of flux.krea using the same method ? Or not yet ?
Fi need to test
SDoes anyone know if I can expect much faster iteration speed (48GB config file) with a H100 rather than a L40S for a full dreambooth finetune ? Thanks
Fa bit faster yes
Fthe full model uses like 28 gb vram
Fso both fits into vram
SThanks 
SIm using this preset configs: "Batch_Size_7_48GB_GPU" -- what GPU on runpod do you reccomend for this: L40s kinda sucks
[DreamBooth Flux Krea]
[DreamBooth Flux Krea]
FA100 faster than L40S
For H100
Fif you want better speed
K@Dr. Furkan Gözükara i was wondering if doing a finetune on krea is a bit different, the woman with teal braids is my finetuned model. Im doing depth to image. workflow i see a lot of realism and shadows on the krea model that gets lost when using my finetuned character krea.
Scan you share the workflow?
Sand i also understand that with classic finetune flux there is no issue?
Kthe depth one or the image to image?
Kim not sure if the classic finetune has that problem, but on krea is really obvious
C@Dr. Furkan Gözükara Dr. Furkan. Do you think that training on the flux krea model in Kohya GUI DreamBooth will be the same as in your tutorial from 22.10.2024:
https://youtu.be/FvpWy1x5etM?si=CI-dObKmDprZesoN
Should I just replace flux.dev with Krea?
Or maybe during this time, some significant optimizations have been introduced that would also be worth changing? Any other options?
Are you planning to do a comparison of flux.dev vs Krea in DreamBooth training?
https://youtu.be/FvpWy1x5etM?si=CI-dObKmDprZesoN
Should I just replace flux.dev with Krea?
Or maybe during this time, some significant optimizations have been introduced that would also be worth changing? Any other options?
Are you planning to do a comparison of flux.dev vs Krea in DreamBooth training?
YouTubeSECourses
If you want to train FLUX with maximum possible quality, this is the tutorial looking for. In this comprehensive tutorial, you will learn how to install Kohya GUI and use it to fully Fine-Tune / DreamBooth FLUX model. After that how to use SwarmUI to compare generated checkpoints / models and find the very best one to generate most amazing image...
JWan Training 
Dis it a bad idea when training a character lora (SDXL) to use two sets of the same images - one in 1024x1024 and one in 896x1152 to be more flexible depending on which aspect ration/resolution i want to create my images with? and if it is not a bad idea, should i create one concept per resolution? would i have to use 2 concepts with reg images?
Dand is there a benefit or downside in using full resolution images (in the correct aspect ratio) or should i downscale everything to the exact resolution (1024x1024/896x1152) or does it downscale on its own when the source images are higher resolution
Fi will test hopefully
Fyes i will do detailed test
Flots of shared vrma
Fit must be very slow
Fostris still doesnt support block swap
Finteresting did he add block swap?
Fit is not as slow as much i would say
Fwhat is your training resolution?
Fi would say make all images exact resolution
Fyou probably can train with 2 different aspect ratio
Fmake sure to enable bucketing
Dthank you. would i need both sets of reg images? so in the end 2 image concepts and 2 reg image concepts? or would you just throw them all in one folder (1 image folder, 1 reg folder)
Fgood question. probably having different aspect ratio would help