Software Engineering Courses (SECourses)•17mo ago

getting this after trying to start. Issue with dir name fixed, new ones come))

novskyOP•8/26/24, 9:17 PM

tried to start rank_5

Nnovsky getting this after trying to start. Issue with dir name fixed, new ones come))

Furkan Gözükara SECourses•8/26/24, 9:22 PM

use dev fp16 model

novskyOP•8/26/24, 9:26 PM

I'm using this one

Nnovsky I'm using this one

Furkan Gözükara SECourses•8/26/24, 9:26 PM

yes it is correct one

Furkan Gözükara SECourses•8/26/24, 9:26 PM

what is your gpu model?

novskyOP•8/26/24, 9:28 PM

maybe t5 is wrong?

Nnovsky maybe t5 is wrong?

Furkan Gözükara SECourses•8/26/24, 9:28 PM

yep it may be wrong :d

novskyOP•8/26/24, 9:28 PM

rtx 4060ti 16gb

Furkan Gözükara SECourses•8/26/24, 9:28 PM

use the downloader bat file i have

novskyOP•8/26/24, 9:30 PM

had to extract only t5 from your python, because downloading flux again will piss me off)

novskyOP•8/26/24, 9:31 PM

thx

Nnovsky had to extract only t5 from your python, because downloading flux again will pis...

Furkan Gözükara SECourses•8/26/24, 9:37 PM

you can edit my files

Furkan Gözükara SECourses•8/26/24, 9:37 PM

remove the ones you dont need

Furkan Gözükara SECourses•8/26/24, 9:37 PM

all my files are editable easy

FFurkan Gözükara SECourses all my files are editable easy

novskyOP•8/26/24, 9:38 PM

yep, just did that, thanks

Xcorpion•8/27/24, 7:59 AM

@Furkan Gözükara SECourses how to add these models to kohya?

Xcorpion•8/27/24, 8:08 AM

just put

kohya_ss\wd14_tagger_model

kohya_ss\wd14_tagger_model

, didnt work

Xcorpion•8/27/24, 8:28 AM

nvm, downloaded another model from kohya and replaced that model with eva02, it works

Xcorpion•8/27/24, 8:28 AM

XXcorpion nvm, downloaded another model from kohya and replaced that model with eva02, it ...

Furkan Gözükara SECourses•8/27/24, 10:47 AM

nice

Evivrus•8/27/24, 10:59 AM

Has anyone received this issue while trying to traing flux with Kohya?

Used the recent single click installer V15 from SECourses Patreon.

Evivrus•8/27/24, 11:00 AM

I know it is probably something to do with accelerate

EEvivrus Has anyone received this issue while trying to traing flux with Kohya? Used th...

Furkan Gözükara SECourses•8/27/24, 11:00 AM

which base model

Furkan Gözükara SECourses•8/27/24, 11:00 AM

you given?

Evivrus•8/27/24, 11:00 AM

the base model form the Blacklabs repo on hugging face

Evivrus•8/27/24, 11:00 AM

from*

EEvivrus the base model form the Blacklabs repo on hugging face

Furkan Gözükara SECourses•8/27/24, 11:01 AM

file name and size please

Furkan Gözükara SECourses•8/27/24, 11:01 AM

also accelerator setup here : https://youtu.be/adVhm9aI9Gc

YouTubeSECourses

Kohya SS GUI How To Set Accelerate Config

Many people having issues due to inaccurate Accelerate setup. This is how to setup to make it work properly

1-click installers and best configs for Kohya FLUX training :

A Big complete tutorial very soon hopefully : https://www.patreon.com/posts/kohya-flux-lora-110293257

Evivrus•8/27/24, 11:01 AM

flux1-dev.safetensors

EEvivrus flux1-dev.safetensors

Furkan Gözükara SECourses•8/27/24, 11:01 AM

ok that is accurate. what is your T5 model name and size?

Evivrus•8/27/24, 11:02 AM

I have tried the combined version: t5xxl_fp16.safetensors
and the non combined version from the blacklabs repo:

Furkan Gözükara SECourses•8/27/24, 11:02 AM

please also apply accelerator setup above and restart and try again

Evivrus•8/27/24, 11:02 AM

EEvivrus Click to see attachment

Furkan Gözükara SECourses•8/27/24, 11:03 AM

this is wrong

Furkan Gözükara SECourses•8/27/24, 11:03 AM

you need to use this one

Furkan Gözükara SECourses•8/27/24, 11:03 AM

comfyanonymous/flux_text_encoders

Evivrus•8/27/24, 11:03 AM

kk will grab and retest

Furkan Gözükara SECourses•8/27/24, 11:03 AM

https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors

Furkan Gözükara SECourses•8/27/24, 11:03 AM

you know i have model downloader too

Evivrus•8/27/24, 11:05 AM

Extra note I do have a 4090 and a 3080 plugged in to the same rig. Though I usually only use 4090 for training 3080 is mainly there for other uses. So I normally set accelerator to only use GPU ID 0