Software Engineering Courses (SECourses)•6mo ago

Anyone have any idea what GPU on MssedCompute is the closest equivalent to RTX 5090's 3s/it for fine

Anyone have any idea what GPU on MssedCompute is the closest equivalent to RTX 5090's 3s/it for finetuning? (single only, not duals). Feel the need for speed :p

Gg7ants80 Getting this now when I start training

Furkan Gözükara SECourses•7/4/25, 11:57 AM

it tells you the error

Furkan Gözükara SECourses•7/4/25, 11:57 AM

did you set your training images folder?

Gghstoic Anyone have any idea what GPU on MssedCompute is the closest equivalent to RTX 5...

Furkan Gözükara SECourses•7/4/25, 11:57 AM

h100 faster

Furkan Gözükara SECourses•7/4/25, 11:57 AM

a100 can be close

ghstoicOP•7/4/25, 11:58 AM

Thanks Doc

diegoarmanda•7/4/25, 12:20 PM

I am currently following and running the training from: https://www.patreon.com/posts/kohya-flux-fine-112099700
Once the training is finished, can the output file be used as a LoRA data file in ComfyUI using the Flux model?

Sorry if my question is confusing, as I am still a beginner.
Thank you!

Patreon

Kohya FLUX Fine Tuning (Full Checkpoints) Training Full Tutorial Fo...

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

FFurkan Gözükara SECourses i recommend after that extract lora

diegoarmanda•7/4/25, 2:35 PM

I am a beginner and would like to ask:
If I train a model using FLUX Full Fine-Tuning or DreamBooth, can the resulting trained file be used as a LoRA model in a ComfyUI workflow?

Sorry if my question is basic, I am still learning. Thank you!

javajunkie•7/4/25, 5:48 PM

Hi, thanks for your excellent tutorials and articles.

I'm having an issue with runpods and LoRA. I completed some tests locally and know have started over with runpods. When I start trainingstart training I get an error you've been asked about before; LoRA type must be set to 'Flux1', 'Flux1 OFT' or 'LyCORIS' if Flux1 checkbox is checked.LoRA type must be set to 'Flux1', 'Flux1 OFT' or 'LyCORIS' if Flux1 checkbox is checked. You've said to another user that they've set the model path incorrect, well my model is in the /workspace/ folder and my path is set as such. I've removed the model and downloaded it fresh, but the error persists, sadly.

~~Am I missing something else?~~

Disregard, loading different base configs seemed to fix it.

Ddiegoarmanda I am a beginner and would like to ask: If I train a model using FLUX Full Fine-T...

Furkan Gözükara SECourses•7/4/25, 11:06 PM

yes

Furkan Gözükara SECourses•7/4/25, 11:06 PM

100%

Jjavajunkie Hi, thanks for your excellent tutorials and articles. ~~I'm having an issue wit...

Furkan Gözükara SECourses•7/4/25, 11:06 PM

great

Badnerle | Can•7/5/25, 12:28 AM

for what reason do we extract lora after dreambooth?

diegoarmanda•7/5/25, 4:31 AM

Hello, I’m still a beginner and learning through SECourses, so I apologize if my question is unclear.

I successfully trained a character using Flux Full Fine Tuning (or DreamBooth), and I tested the result in ComfyUI as a UNet model (without using flux1.Dev). The results were very realistic and successful.

Now I’m wondering:

Can this trained model be used as a replacement for Flux Kontent?

Or, can I use Flux Kontent and combine it with my trained model as a LoRA?

Thank you for your help!

BBadnerle | Can for what reason do we extract lora after dreambooth?

Furkan Gözükara SECourses•7/5/25, 11:52 AM

if you need lora

Furkan Gözükara SECourses•7/5/25, 11:52 AM

otherwise you dont need and i dont recommned dreambooth works better

Ddiegoarmanda Hello, I’m still a beginner and learning through SECourses, so I apologize if my...

Furkan Gözükara SECourses•7/5/25, 11:52 AM

1 : nope different model

Furkan Gözükara SECourses•7/5/25, 11:52 AM

2 : nope sadly

Furkan Gözükara SECourses•7/5/25, 11:53 AM

but i didnt test lora on kontext yet so cant say for 100%

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭•7/6/25, 1:04 PM

I want to fine tune on a specific quality type of images of women. Not 1 character, but I want the fine tune to be able to generate woman with same quality as training data. Say I have pics of 5 to 10 different women of multiple shapes, hair, body, ethnicity etc. How many pics would be needed for a high quality finetune? Ideally it should be able to generate women not in the training set, I.e. an Arab woman, but with same quality as training pics. Unsure how to proceed, only done fine tune with 1 character so far. I guess captions should be needed here.

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭I want to fine tune on a specific quality type of images of women. Not 1 charact...

Furkan Gözükara SECourses•7/6/25, 10:56 PM

you may try bigger fine tunes people published with good prompting

Furkan Gözükara SECourses•7/6/25, 10:56 PM

otherwise you are looking for rather general fine tune

Furkan Gözükara SECourses•7/6/25, 10:57 PM

so you need more images the better and use good captions

Henry•7/7/25, 7:20 PM

Hello! I am trying to get the kaggle notebook to run "kohya-sdxl-lora-training-on-a-free-kaggle-notebook_v28", but the moment I import it just starts downloading a ton of dependencies automatically. I did not see this in the tutorial. What is the expected behavior? Thanks in advance

Henry•7/7/25, 7:24 PM

Also, I can seem to edit the GPUs or the folder directories to point to training images

FFurkan Gözükara SECourses yes ai video upscaler is my priority

Cryptozow.x•7/7/25, 9:46 PM

Hey, just wanted to throw this here FYI in case you're looking for STAR alternatives within your APP, you can check if you want, they claim to be better than STAR, however, I haven't been able to test it out.
https://github.com/yjsunnn/DLoRAL?tab=readme-ov-file

GitHub

GitHub - yjsunnn/DLoRAL: Official Implementation of “One-Step Dif...

Official Implementation of “One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution” - yjsunnn/DLoRAL

HHenry Hello! I am trying to get the kaggle notebook to run "kohya-sdxl-lora-training-o...

Furkan Gözükara SECourses•7/7/25, 10:09 PM

yes we had to update a lot

HHenry Also, I can seem to edit the GPUs or the folder directories to point to training...

Furkan Gözükara SECourses•7/7/25, 10:10 PM

yes you have to edit them

CCryptozow.x Hey, just wanted to throw this here FYI in case you're looking for STAR alternat...

Furkan Gözükara SECourses•7/7/25, 10:10 PM

i just checked

Furkan Gözükara SECourses•7/7/25, 10:10 PM

looks weak

Henry•7/7/25, 10:10 PM

I got it to work now!

HHenry I got it to work now!

Furkan Gözükara SECourses•7/7/25, 10:11 PM

great

underdog•7/9/25, 12:25 AM

Hello,

I'm using the configuration you have been using for 48 gb ram. I rented an H100 GPU with 20 image dataset but the images generated after training does not even match with my training data. Where do you think I might be going wrong? all the parameters are loaded from 15 image dataset config file. Can you please advise?

Uunderdog Hello, I'm using the configuration you have been using for 48 gb ram. I rented ...

Furkan Gözükara SECourses•7/9/25, 12:29 AM

which config you used which zip file

Furkan Gözükara SECourses•7/9/25, 12:29 AM

please more info

FFurkan Gözükara SECourses which config you used which zip file

underdog•7/9/25, 12:30 AM

I used 48GB_GPU_28200MB_6.3_second_it_Tier_1.json

Uunderdog I used 48GB_GPU_28200MB_6.3_second_it_Tier_1.json

Furkan Gözükara SECourses•7/9/25, 12:31 AM

ok it is a dreambooth config. did you load into dreambooth tab?

Furkan Gözükara SECourses•7/9/25, 12:31 AM

properly did set training images folder?

Furkan Gözükara SECourses•7/9/25, 12:32 AM

also we have an example dataset

Furkan Gözükara SECourses•7/9/25, 12:32 AM

https://www.patreon.com/posts/114972274

Patreon

Perfect Quality Example Model Training Images Dataset - Can Be Used...

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

underdog•7/9/25, 12:33 AM

yes I'm loading them in the dream booth tab. I also am loading the training images properly but just to be sure is there a way for me to know if the training images are being loaded properly or not?

underdog•7/9/25, 12:35 AM

I can see the folder being created after clicking prepare training data

Uunderdog I can see the folder being created after clicking prepare training data

Furkan Gözükara SECourses•7/9/25, 12:36 AM

ok this is accurate

Furkan Gözükara SECourses•7/9/25, 12:36 AM

you can look at the cmd logs

Furkan Gözükara SECourses•7/9/25, 12:36 AM

to see

Furkan Gözükara SECourses•7/9/25, 12:36 AM

if you copy paste here i can understand

underdog•7/9/25, 12:39 AM

I get this when I press prepare training data

/kaggle/working/kohya_ss/training_data/img/1_ohwx
man...
00:37:48-667704 INFO Regularization images directory is missing... not
copying regularisation images...
00:37:48-670635 INFO Done creating kohya_ss training folder structure at

and when I start training, you can see them in the message.txt file.

Thanks a lot for your help.

message.txt546.13KB

underdog•7/9/25, 12:51 AM

I tried other images including mine but didn't work, so I tried taking a few screenshots from the 15 image dataset of yours and at the 1000th step the sample image I get is not even close to the training images. I will try with Dwayne Johnson images now and give you an update.

gFspxwNfsQP3re7IZvx8hk0A5lXeyp8hgGgdEKqffwAxUyyrgpMhgAAAABJRU5ErkJggg.png

Uunderdog I get this when I press prepare training data /kaggle/working/kohya_ss/trainin...

Furkan Gözükara SECourses•7/9/25, 12:52 AM

last time i tested

Furkan Gözükara SECourses•7/9/25, 12:52 AM

kaggle was not able to train flux

Furkan Gözükara SECourses•7/9/25, 12:53 AM

"/kaggle/working/kohya_ss/traini
ng_data/img/1_ohwx man"

Anyone have any idea what GPU on MssedCompute is the closest equivalent to RTX 5090's 3s/it for fine

Similar Threads