Software Engineering Courses (SECourses)•17mo ago

You just gave me an idea to see if Flux can understand hex codes from the prompt:

�

𝔸𝖨ΞӾOP•8/28/24, 1:23 PM

Looks like it does have some kinda effect lol

Ddsienra can you do a fix .bat to not have to reinstall kohya completely?

Furkan Gözükara SECourses•8/28/24, 1:23 PM

yep

Furkan Gözükara SECourses•8/28/24, 1:23 PM

let me upload

Furkan Gözükara SECourses•8/28/24, 1:23 PM

dropped from 8 second to 5.8

�

𝔸𝖨ΞӾOP•8/28/24, 1:24 PM

Oh btw I use https://fastflux.ai/ when I quickly want to test out ideas for prompts

FastFLUX | Instant FLUX Image Creation for Free

Create beautiful FLUX images in milliseconds with FastFLUX. Free, fast, and no sign-up required. Image generation powered by Runware.

�𝔸𝖨ΞӾ You just gave me an idea to see if Flux can understand hex codes from the prompt...

umuthasanoglu•8/28/24, 1:24 PM

Wow. I tried on SD models it did not work but haven't tried Flux yet. Good job. I think it's doing good but can you try more?

Uumuthasanoglu Wow. I tried on SD models it did not work but haven't tried Flux yet. Good job. ...

�

𝔸𝖨ΞӾOP•8/28/24, 1:25 PM

I used https://coolors.co/ for the palettes

Coolors.co

Coolors - The super fast color palettes generator!

Generate or browse beautiful color combinations for your designs.

umuthasanoglu•8/28/24, 1:25 PM

We may even create wildcards and/or lora's with this

Uumuthasanoglu We may even create wildcards and/or lora's with this

�

𝔸𝖨ΞӾOP•8/28/24, 1:28 PM

Hmm, I think perhaps using IP Adapter with just a color palette may work better tho. Haven't tested it out yet

umuthasanoglu•8/28/24, 1:29 PM

Yes but I want an easier way. Maybe it could be color palette presets. I could define color palette with common names + hex codes for each color style

umuthasanoglu•8/28/24, 1:32 PM

In illustrations, it's really important to define key color palettes to have a unique and constant art style

Furkan Gözükara SECourses•8/28/24, 1:35 PM

https://www.patreon.com/posts/110879657

Patreon

Kohya FLUX LoRA Training Full Tutorial For Local Windows and Cloud ...

Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon

Furkan Gözükara SECourses•8/28/24, 1:35 PM

28 August 2024 Update
340 prompts txt added to the testing prompts list

Install_Torch_2_5_Dev_Huge_Speed_Up.bat file added

This will install nightly development branch of Torch

It brings huge Speed up on Windows and you can get closer speeds to do training in Linux

I haven't tested its quality yet so it is experimental

Furkan Gözükara SECourses•8/28/24, 1:40 PM

i am training right now to verify its quality on my rtx 3060

FFurkan Gözükara SECourses Click to see attachment

smoke•8/28/24, 1:59 PM

Is this the recent clip_l trained one?

smoke•8/28/24, 1:59 PM

And are you using captions for these?

Ssmoke Is this the recent clip_l trained one?

Furkan Gözükara SECourses•8/28/24, 2:01 PM

only ohwx man atm

Furkan Gözükara SECourses•8/28/24, 2:02 PM

yes clip 5e-05

FFurkan Gözükara SECourses but nothing good as supir

pat._.fenis•8/28/24, 2:12 PM

Have there been any updates that make Supir useable for video?
When using settings 1,1,2,1, a batch upscale of 8700 images with the original resolution at 960x576 and using Upscale Size 1 it still takes nearly 16 hours using a 4090...
If this is how Supir is supposed to function it is not a viable option for video.

FFurkan Gözükara SECourses only ohwx man atm

axel•8/28/24, 2:12 PM

I send you message in private, I think to be important especially for lora training

Ppat._.fenis Have there been any updates that make Supir useable for video? When using settin...

Furkan Gözükara SECourses•8/28/24, 2:19 PM

they are working on video model

pat._.fenis•8/28/24, 2:20 PM

I am batch upscaling individual images that comprise video, have there been any updates that speed up Supir for batch processing?

pat._.fenis•8/28/24, 2:21 PM

I just don't understand why it is so painfully slow...

Maxivi•8/28/24, 2:34 PM

@Dr. Furkan Gözükara i think your kohya installer is broken.

im now getting windows os errors, and gui itself doesn't want to open.

MMaxivi @Dr. Furkan Gözükara i think your kohya installer is broken. im now getting win...

Furkan Gözükara SECourses•8/28/24, 2:40 PM

can you tell me error

Ppat._.fenis I am batch upscaling individual images that comprise video, have there been any ...

Furkan Gözükara SECourses•8/28/24, 2:40 PM

sadly not but i feel you

Furkan Gözükara SECourses•8/28/24, 2:41 PM

i wonder if pytorch 2.5 can speed it up

pat._.fenis•8/28/24, 2:43 PM

It shows I am only using 7.2gb of my GPU ram

pat._.fenis•8/28/24, 2:44 PM

also states gradio 4.29.0 is available and it is using 4.19 during startup selection

FFurkan Gözükara SECourses can you tell me error

Maxivi•8/28/24, 2:44 PM

well im not getting the win os error now because the gui.bat is not doing anything anymore

Maxivi•8/28/24, 2:45 PM

its just black screen

Maxivi•8/28/24, 2:45 PM

i was getting permission errors as well

MMaxivi well im not getting the win os error now because the gui.bat is not doing anythi...

Furkan Gözükara SECourses•8/28/24, 2:46 PM

are you on windows 11?

Maxivi•8/28/24, 2:46 PM

yessir

Furkan Gözükara SECourses•8/28/24, 2:47 PM

windows 11 sucks

Furkan Gözükara SECourses•8/28/24, 2:47 PM

run gui.bat inside kohya ss folder and tell me waht you see

Furkan Gözükara SECourses•8/28/24, 2:47 PM

restart computer and try again too

Maxivi•8/28/24, 2:47 PM

i know i had to upgrade though, windows 10 was giving so much issues, random pc shutdowns, slowdowns etc. plus windows 10 soon wont be supported anymore

FFurkan Gözükara SECourses restart computer and try again too

Maxivi•8/28/24, 2:48 PM

already tried i was getting just a blank terminal

Maxivi•8/28/24, 2:48 PM

but im doing system restore right now back to 4 days ago

Furkan Gözükara SECourses•8/28/24, 2:48 PM

did you get any error at install?

Furkan Gözükara SECourses•8/28/24, 2:48 PM

windows 11 has problematic permissions with python apps :/

Maxivi•8/28/24, 2:48 PM

yeah i was getting some windows os error

MMaxivi yeah i was getting some windows os error

Furkan Gözükara SECourses•8/28/24, 2:48 PM

you need to fix it

Furkan Gözükara SECourses•8/28/24, 2:48 PM

it is permission related error

Maxivi•8/28/24, 2:49 PM

https://tenor.com/view/sad-gif-19231393

Tenor

Furkan Gözükara SECourses•8/28/24, 2:50 PM

i still use windows 10

but i also get random shutdowns and bluescreens sometimes :/

pixelscience•8/28/24, 2:52 PM

Hey there, thank you for the plethora of information, I'm learning a ton from all the work you're putting into this.

I'm trying to understand how Koyha is calculating the total optimization steps when I run my LORA training...

I have 30 images set to 10 repeats, so that part is clear, but it somehow seems to be automatically configuring things to hit 1600 steps.

running training / 学習開始
  num train images * repeats / 学習画像の数×繰り返し回数: 300
  num reg images / 正則化画像の数: 0
  num batches per epoch / 1epochのバッチ数: 300
  num epochs / epoch数: 6
  batch size per device / バッチサイズ: 1
  gradient accumulation steps / 勾配を合計するステップ数 = 1
  total optimization steps / 学習ステップ数: 1600

running training / 学習開始
  num train images * repeats / 学習画像の数×繰り返し回数: 300
  num reg images / 正則化画像の数: 0
  num batches per epoch / 1epochのバッチ数: 300
  num epochs / epoch数: 6
  batch size per device / バッチサイズ: 1
  gradient accumulation steps / 勾配を合計するステップ数 = 1
  total optimization steps / 学習ステップ数: 1600

The Epoch setting in Koyha doesn't seem to actually change anything it looks like, so where does it come up with 6 epochs?

dsienra•8/28/24, 3:06 PM

I'm doing new tests with TE training multiple people but it seems that still bleeds to the class, may be for flux you have to remove the class completely to avoid bleeding the people you are training between each other, and the bleed to the entire class "man" for example, may be changing the captions from "ohwx man" to only "ohwx" will do the magic, flux already knows what a man is and is not necessary to add it to the captions, In sdxl this wasn't a problem for me. the model learned that "name1 man" was different from "name2 man" in flux I didn't find a solution yet, I will try removing the class in my next training.

Ppixelscience Hey there, thank you for the plethora of information, I'm learning a ton from al...

Furkan Gözükara SECourses•8/28/24, 3:06 PM

i explained in the video

running training / 学習開始 num train images * repeats / 学習画像の数×繰り返し回数: 300 num reg images / 正則化画像の数: 0 num batches per epoch / 1epochのバッチ数: 300 num epochs / epoch数: 6 batch size per device / バッチサイズ: 1 gradient accumulation steps / 勾配を合計するステップ数 = 1 total optimization steps / 学習ステップ数: 1600

You just gave me an idea to see if Flux can understand hex codes from the prompt:

Similar Threads

Similar Threads

Similar Threads