Software Engineering Courses (SECourses)•11mo ago

@Furkan Gözükara SECourses in video you say epoch 100 is good for 256 images... but after you menti

@Furkan Gözükara SECourses
in video you say epoch 100 is good for 256 images... but after you mention that epoch 50 is best?

what I dont understand yet... so while training you can already move the generated checkpoints to some different place because these are kind of finished states of the trained model and will only be needed for training in case the training fails. which would mean you could then use the checkpoint as new starting point? is this correct?

1nfern0•2/9/25, 5:24 PM

I've read some comments about using a clip interogator to filter out similar images from a dataset. Has anyone used this technique or know of a tool that could do this?

Vexmachina•2/9/25, 5:39 PM

Is anyone aware of a platform that offers the ability to train a flux lora and then use something like adetailer/udetailer to fix the face on inference, like out of the box or with minimal setup?

Not looking to set up my own thing from scratch on something like runpod, hoping for something more plug and play like Replicate

AIGambino•2/9/25, 8:09 PM

https://xh9998.github.io/DiffVSR-project/

DiffVSR

..tarkan Little comparison to see the improvement side by side.

Macsalim•2/9/25, 9:17 PM

Thank you for the Flux workflow, it looks promising. But I have a question I keep getting a "mat1 mat2 mismatch" error for Clip_G looks like I'm using the wrong version, can you share the link to the clip file you're using?
Also, it would be nice if you can share the generation data for one of your images, just for me to check if everything is working correctly.

MMacsalim Thank you for the Flux workflow, it looks promising. But I have a question I kee...

.tarkan•2/9/25, 10:19 PM

Here the clip g model. https://huggingface.co/stabilityai/stable-diffusion-3.5-large/resolve/main/text_encoders/clip_g.safetensors?download=true

Mido•2/10/25, 2:52 AM

Hello

Mido•2/10/25, 2:53 AM

Whenever i train kohyass lora on a flux model, i get error "returned non-zero exit status 1."

Mido•2/10/25, 2:53 AM

other models it works perfectly fine

Aragon•2/10/25, 12:59 PM

I'm getting two additional options for Kohya that don't appear on the exapmle image on the guide on Patreon - "Conv Dimension" and "Minimum Difference". Should I leave them with default values?

VVioletAnt ahhh ok I think I understand. more images higher quality and less epochs needed

Furkan Gözükara SECourses•2/10/25, 1:52 PM

yes batch size 1 is best

AAiInfluence @Furkan Gözükara SECourses SOS 🚨🚨 I need suggestions for good lip sync

Furkan Gözükara SECourses•2/10/25, 1:53 PM

i dont know good one atm

VVioletAnt @Furkan Gözükara SECourses in video you say epoch 100 is good for 256 images......

Furkan Gözükara SECourses•2/10/25, 1:53 PM

the logic is like this

Furkan Gözükara SECourses•2/10/25, 1:53 PM

we do up to 200 epochs

Furkan Gözükara SECourses•2/10/25, 1:53 PM

and we compare and find best one

Furkan Gözükara SECourses•2/10/25, 1:53 PM

but if you have too many images

Furkan Gözükara SECourses•2/10/25, 1:53 PM

it is becoming redundant

11nfern0 I've read some comments about using a clip interogator to filter out similar ima...

Furkan Gözükara SECourses•2/10/25, 1:53 PM

we have image similarity app you can try

VVexmachina Is anyone aware of a platform that offers the ability to train a flux lora and t...

Furkan Gözükara SECourses•2/10/25, 1:54 PM

our massed compute image has swarmui

Furkan Gözükara SECourses•2/10/25, 1:55 PM

just click updat first

Furkan Gözükara SECourses•2/10/25, 1:55 PM

then ready

AAIGambino https://xh9998.github.io/DiffVSR-project/

Furkan Gözükara SECourses•2/10/25, 1:55 PM

github not found https://github.com/xh9998/DiffVSR

MMido Whenever i train kohyass lora on a flux model, i get error "returned non-zero e...

Furkan Gözükara SECourses•2/10/25, 1:56 PM

hard to know without seeing everything

AAragon I'm getting two additional options for Kohya that don't appear on the exapmle im...

Furkan Gözükara SECourses•2/10/25, 1:56 PM

you want to extract flux lora?

Furkan Gözükara SECourses•2/10/25, 1:56 PM

that is inaccurate tab

Furkan Gözükara SECourses•2/10/25, 1:56 PM

click this

Furkan Gözükara SECourses•2/10/25, 1:56 PM

FFurkan Gözükara SECourses click this

Aragon•2/10/25, 1:57 PM

Oops, my bad

. Thanks

AiInfluence•2/10/25, 1:59 PM

@Furkan Gözükara SECourses do you think to have 2 Lora people in one picture it's best to do face inpainting or regional prompting on Swarm? what will achieve best results

AAiInfluence @Furkan Gözükara SECourses do you think to have 2 Lora people in one picture it'...

Furkan Gözükara SECourses•2/10/25, 2:00 PM

probably face inpainting

FFurkan Gözükara SECourses probably face inpainting

AiInfluence•2/10/25, 2:01 PM

with yolo segmentation or individual swarm masking using flux inpaint model?

FFurkan Gözükara SECourses hard to know without seeing everything

Mido•2/10/25, 2:14 PM

Mido•2/10/25, 2:15 PM

last_20250210-092037.json5.39KB

config_lora-20250210-092037.toml1.51KB

VioletAntOP•2/10/25, 3:30 PM

@Furkan Gözükara SECourses finetuning can or cannot use 2 gpu? Not used in my case. Are there any settings?

VVioletAnt @Furkan Gözükara SECourses finetuning can or cannot use 2 gpu? Not used in my ca...

Aragon•2/10/25, 3:40 PM

dreambooth-training

VVioletAnt @Furkan Gözükara SECourses finetuning can or cannot use 2 gpu? Not used in my ca...

AiInfluence•2/10/25, 5:30 PM

it can't

julius•2/10/25, 8:53 PM

you guys ever get this error while training? in this session im running now, seems to happen almost every time, just have to click "generate" a few times and then it works anyway?
(4x RTX 6000 on MassedCompute, using Init Image)

Pelolisu•2/10/25, 10:11 PM

Hi guys, I made this little toy for you: https://huggingface.co/spaces/elismasilva/mixture-of-diffusers-sdxl-tiling

Mixture Of Diffusers SDXL Tiling - a Hugging Face Space by elismasilva

m3dia_offline•2/10/25, 10:40 PM

@Furkan Gözükara SECourses have you tried the recently released flux pro model's api for fine tuning, wondering what your take is on that vs the dreambooth method in terms of quality

PPelolisu Hi guys, I made this little toy for you: https://huggingface.co/spaces/elismasil...

Furkan Gözükara SECourses•2/10/25, 11:07 PM

amazing

Mm3dia_offline @Furkan Gözükara SECourses have you tried the recently released flux pro model's...

Furkan Gözükara SECourses•2/10/25, 11:07 PM

people tested and it is inferior

FFurkan Gözükara SECourses people tested and it is inferior

m3dia_offline•2/10/25, 11:08 PM

i see, powering through your nearly 3 hour dreambooth tutorial on flux since yesterday haha

Furkan Gözükara SECourses•2/11/25, 12:06 AM

sana controlnet published

Furkan Gözükara SECourses•2/11/25, 12:06 AM

i should implement

Dazzastrous [4090]•2/11/25, 12:06 AM

nice

Furkan Gözükara SECourses•2/11/25, 12:08 AM

replied all emails and messages time to sleep

VioletAntOP•2/11/25, 1:15 AM

@Furkan Gözükara SECoursesEDIT SOLVED: In the folder on massed compute there was hidden folder "cache" that I had to delete. then I could upload again. ORIGINAL MESSEAGE: I am trying to transfer checkpoints from massed compute to huggingface. I use the jupyter notebook. Works very good and fast. But I have one problem.... Some checkpoints I had to delete from hugging face. Then I tried again to retransfer them from massed compute to hugging face. but the jupyter always says: "Recovering from metadata files: 100%". and then its just skipping.... What do I need to do? is there an overwrite command? or do i need to delete metadata files and where are they?

FFurkan Gözükara SECourses people tested and it is inferior

m3dia_offline•2/11/25, 1:24 AM

do you think a dreambooth or lora training of just text can create a reliable text generating model?

.tarkan•2/11/25, 8:21 AM

@Furkan Gözükara SECourses if I set the guidance scale to 6 in the kohya fine tune parameters, will that influence training and if yes how?

kareem9000•2/11/25, 10:47 AM

kareem9000•2/11/25, 10:47 AM

anyone else get python crashing when they load swarmui on the first time? running it locally

@Furkan Gözükara SECourses in video you say epoch 100 is good for 256 images... but after you menti

Similar Threads