Software Engineering Courses (SECourses)•3y ago

am trying to train with kohya script and i i can only make it get this far. is there any help to adv

am trying to train with kohya script and i i can only make it get this far. is there any help to advise here? it'll stay at this step all night and do nothing if i leave it.

  total optimization steps / 学習ステップ数: 4350

  total optimization steps / 学習ステップ数: 4350

steps:   0%|                                                                                  | 0/4350 [00:00<?, ?it/s]

steps:   0%|                                                                                  | 0/4350 [00:00<?, ?it/s]

epoch 1/15

epoch 1/15

Fflowwolf anybody got any ideas? i'm trying to use kohya for the first time and i've got i...

Furkan Gözükara SECourses•5/25/23, 9:48 PM

try adamw

Furkan Gözükara SECourses•5/25/23, 9:48 PM

not 8bit

Fflowwolf am trying to train with kohya script and i i can only make it get this far. is t...

Furkan Gözükara SECourses•5/25/23, 9:48 PM

do you get any error?

Furkan Gözükara SECourses•5/25/23, 9:48 PM

looks like 0 steps

flowwolfOP•5/25/23, 9:48 PM

none. it'll just stay there

flowwolfOP•5/25/23, 9:49 PM

i've had errors whiel i was building the data set and i learned to structure those folders right and got through it. this is a brick wall though. an error would aid. haha. trying adamw now

NNull does anyone know how much wound it cost me if I use stablediffusion finetuning (...

Furkan Gözükara SECourses•5/25/23, 9:49 PM

i dont know them but runpod pretty cheap

flowwolfOP•5/25/23, 9:51 PM

yeah same thing again. it fires up really fast and efficient. detects my image set. begins the training and just halts up. another thing i've noticed is if it's there and i press train in the gui again, it'll run again and get to that point. /shrug

Fflowwolf yeah same thing again. it fires up really fast and efficient. detects my image s...

Furkan Gözükara SECourses•5/25/23, 9:52 PM

very interesting

Furkan Gözükara SECourses•5/25/23, 9:52 PM

did you try with torch 1.13?

flowwolfOP•5/25/23, 9:55 PM

just the default 1.12.1 with the cudnn fixes since i have a 4080. hmm will try 1.13

flowwolfOP•5/25/23, 9:55 PM

i tried the experimental 2 but ran into some startup issues

Ffelixsanz Click to see attachment

Null•5/25/23, 9:58 PM

where to download pix2pix?

FFurkan Gözükara SECourses i think now it works . they made updates

majidzendaniyan•5/25/23, 9:58 PM

Which one should I download?

majidzendaniyan•5/25/23, 9:58 PM

I had downloaded the first one in the past

Mmajidzendaniyan Which one should I download?

Null•5/25/23, 9:59 PM

big one if your gpu can fit

NNull where to download pix2pix?

felixsanz•5/25/23, 10:01 PM

google

Ffelixsanz google

Null•5/25/23, 10:01 PM

are you using pruned?

felixsanz•5/25/23, 10:01 PM

pruned what

Null•5/25/23, 10:01 PM

pruned pix2pix

Null•5/25/23, 10:01 PM

or the 7.7GB one

felixsanz•5/25/23, 10:02 PM

i don't use pix2pix, and if you check my picture you would see the model i was using

PPickledJoi if you are running auto locally you aren't using the internet. your machine is p...

Lacerda613•5/25/23, 10:02 PM

It will not start if I disable my Wi-Fi.

Fflowwolf ui-config.json in automatic1111 root directory allows the change. i don't think...

Lacerda613•5/25/23, 10:03 PM

Thank you.

flowwolfOP•5/25/23, 10:22 PM

i did it! i got 1.13.1+cu117 to work! also copied the cudnn dlls into the torch/lib folder for good measure. it didn't work again, but there was a message that showed up this time! Right at the start of training before the rest of the steps like image set detection occur.

[WinError 127] The specified procedure could not be found

[WinError 127] The specified procedure could not be found

WARNING:root:WARNING: [WinError 127] The specified procedure could not be found

WARNING:root:WARNING: [WinError 127] The specified procedure could not be found

Need to compile C++ extensions to get sparse attention suport. Please run python setup.py build develop

Need to compile C++ extensions to get sparse attention suport. Please run python setup.py build develop

prepare tokenizer

prepare tokenizer

Fflowwolf i did it! i got 1.13.1+cu117 to work! also copied the cudnn dlls into the torch...

Furkan Gözükara SECourses•5/25/23, 10:35 PM

but working now?

flowwolfOP•5/25/23, 10:36 PM

nope!

i dont know what else to do . i think i broke it by trying to install new xformers. might just keep experimenting in the webui-extension but i can't seem to get styles to not look over trained there.

flowwolfOP•5/25/23, 10:38 PM

no indication of what the problem is is really quite frustrating. not sure why the script won't log anything or error out or what

felixsanz•5/25/23, 10:43 PM

when comparing checkpoints (steps, methods, etc), what parameters do you guys use on XYZ? only seed and checkpoint?

flowwolfOP•5/25/23, 10:45 PM

i do batches of 4 so each cell gets 4 seeds anyways. checkpoint and cfg are ones i'll try. sometimes prompt s/r. you can try different prompts with it or different lora weights. the :1> can s/r to :0.5>, :1.5>, etc. more a lora thing

felixsanz•5/25/23, 10:46 PM

what cfgs you try? and why?

Ffelixsanz when comparing checkpoints (steps, methods, etc), what parameters do you guys us...

Furkan Gözükara SECourses•5/25/23, 10:47 PM

you can try cfg

Furkan Gözükara SECourses•5/25/23, 10:47 PM

and perhaps pre selected seeds

Furkan Gözükara SECourses•5/25/23, 10:47 PM

added some more prompts : https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Midjourney-Level-Realism.md

GitHub

Stable-Diffusion/Midjourney-Level-Realism.md at main · FurkanGozuka...

Best Stable Diffusion and AI Tutorials, Guides, News, Tips and Tricks - Stable-Diffusion/Midjourney-Level-Realism.md at main · FurkanGozukara/Stable-Diffusion

flowwolfOP•5/25/23, 10:48 PM

5-9[5] is my usual setting. when i'm generating its around 7 usually but if im testing a model i want to see what a range looks like

felixsanz•5/25/23, 10:48 PM

pre selected seeds = try seeds until you have some that look great, so you can compare on best possible seeds?

Ffelixsanz pre selected seeds = try seeds until you have some that look great, so you can c...

Furkan Gözükara SECourses•5/25/23, 10:48 PM

yes

Fflowwolf nope! i dont know what else to do . i think i broke it by trying to install ne...

Furkan Gözükara SECourses•5/25/23, 10:48 PM

if you are our patreon supporter i can connect your computer and install

flowwolfOP•5/25/23, 10:48 PM

maybe sometime thanks

FFurkan Gözükara SECourses you can try cfg

felixsanz•5/25/23, 10:49 PM

what cfgs do you try?

felixsanz•5/25/23, 10:49 PM

4,7,11 is fine?

Ffelixsanz 4,7,11 is fine?

Furkan Gözükara SECourses•5/25/23, 10:50 PM

i try usually 4,5,6,7,8,9,10,11,12

Furkan Gözükara SECourses•5/25/23, 10:50 PM

if i do

felixsanz•5/25/23, 10:51 PM

but at the same time? X=seed, Y=checkpoint, Z=cfg?

Ffelixsanz but at the same time? X=seed, Y=checkpoint, Z=cfg?

Furkan Gözükara SECourses•5/25/23, 10:51 PM

yes you can use

Furkan Gözükara SECourses•5/25/23, 10:51 PM

but which order looks best

Furkan Gözükara SECourses•5/25/23, 10:51 PM

you need to test

felixsanz•5/25/23, 10:52 PM

i'm gonna use seed as Z i think, as it's separated in "another dimension"

felixsanz•5/25/23, 10:55 PM

let's see if the freaking plot script is not still broken... Xd 8640 steps

LLacerda613 It will not start if I disable my Wi-Fi.

PickledJoi•5/25/23, 10:55 PM

it should start without a problem if you already installed the dependencies (by launching with your internet on at least once). what exactly are you doing to disable your wifi

am trying to train with kohya script and i i can only make it get this far. is there any help to adv

Similar Threads

Similar Threads

Similar Threads