Software Engineering Courses (SECourses)•15mo ago

@Dr. Furkan Gözükara time for you to stay away and run more tests. lycoris training available now f

@Dr. Furkan Gözükara time for you to stay away and run more tests.

lycoris training available now for kohya and he also fixed the prior preservation loss

MaxiviOP•10/17/24, 11:51 PM

he will merge into the branch in a bit

MMaxivi @Dr. Furkan Gözükara time for you to stay away and run more tests. lycoris trai...

Furkan Gözükara SECourses•10/17/24, 11:58 PM

i will more tests later i need to publish tutorial first

Furkan Gözükara SECourses•10/17/24, 11:58 PM

rest will be only updating configs

MaxiviOP•10/17/24, 11:59 PM

Lycoris already looks quite good with dora on flux, so hopefully you can explore better settings when you do

FFurkan Gözükara SECourses just wait till you try fine tuning

xlads•10/18/24, 12:20 AM

I can’t wait for this ahahah + can wait for 4K image to video!

BrennenRB•10/18/24, 12:33 AM

Hello all- I am looking for help with a project relating to flux and creating seamless images. Right now, it is unable to make images that can be tiled into larger designs. If anyone is up for a project, I am willing to pay for its development.

AIGambino•10/18/24, 1:31 AM

dr for the style, what happens if you add in like I dont know 5,000 images for the stylization then you do half that number for your character. Is that too much?

dsienra•10/18/24, 1:47 AM

https://huggingface.co/nyanko7/flux-dev-de-distill Unet LR 0.0001 vs 0.00003 the loos curve looks so different, up 80 epochs at 0.0001 it started to overtrain, lets see now...

nyanko7/flux-dev-de-distill · Hugging Face

dsienra•10/18/24, 1:52 AM

may be 0.00003 is too low, idk

Ddsienra may be 0.00003 is too low, idk

druhl•10/18/24, 2:03 AM

Isn't this like, already 7.5 times higher than what the Doc recommended?

dsienra•10/18/24, 2:04 AM

Ddruhl Isn't this like, already 7.5 times higher than what the Doc recommended?

dsienra•10/18/24, 2:09 AM

this is one of the Dr. Furkan config files 3e-05=0.00003

dsienra•10/18/24, 2:12 AM

that's with the T5 options enable, may be I have to increase it to 0.00005 in this file is at that value

druhl•10/18/24, 2:13 AM

Ohh, I see. You're doing a lora, not a full dreambooth. My bad.

dsienra•10/18/24, 2:13 AM

yes lora

dsienra•10/18/24, 2:15 AM

I will test a finetune in the future with flux-dev-de-distill but it uses cfg scale 3.5 that is slower, I can try to extract the lora and try it with regular flux-dev at full speed

druhl•10/18/24, 2:16 AM

Yes, makes sense

Pelolisu•10/18/24, 2:24 AM

Testing Style transfer function in new ui, is being born.

FFurkan Gözükara SECourses i did with onetrainer

deepblue_1337•10/18/24, 4:22 AM

and do you recommend masked training over "normal" training?

deepblue_1337•10/18/24, 4:24 AM

also as far as I remember your dataset is images without a detailed background anyway, right?

Nick44•10/18/24, 4:49 AM

just had amazing results training a flux fine tune , thank u

NNick44 just had amazing results training a flux fine tune , thank u

Nick44•10/18/24, 5:05 AM

also how do I switch fp16 models in flux without this message

BBrennenRB Hello all- I am looking for help with a project relating to flux and creating se...

xlads•10/18/24, 5:59 AM

Hello, sent you a request

xlads•10/18/24, 6:00 AM

@BrennenRB

PPelolisu Testing Style transfer function in new ui, is being born.

xlads•10/18/24, 6:00 AM

Interesting!

Jackie•10/18/24, 6:02 AM

I have a dataset of 25 images for this character, and I’m trying to train a LoRA for FLUX. I used the best config with Rank 4 and set the epoch to 100, which resulted in 2500 steps. However, the results from the prompt don’t look like the character I want to portray. Should I increase the epoch to 200? Also, I’m using captions generated by ChatGPT like this: 'GDZHBo_flx, a cartoon character with a large mustache and glasses, sitting with a playful expression as he appears to be gesturing something with his hands, adding a humorous and relaxed vibe.' Any advice?

Jackie•10/18/24, 6:07 AM

This is an example of the results I’m getting after training the character. The outputs don’t quite match the character I want

Ddruhl Can I have SwarmUI find models in my comfy unet folder? lol

druhl•10/18/24, 6:55 AM

This works (Saved me tons of space on home PC):

You should be able to change ModelRoot to whatever place on your system you have your models stored and, if the subdirs vary from what SwarmUI uses, modify them here. That should capture unet. It will look in {ModelRoot}/unet or {ModelRoot}/diffusion_models as needed.