Software Engineering Courses (SECourses)•17mo ago

I've wondered about the avr_loss and tensorboard graphs available during training... anything to loo

I've wondered about the avr_loss and tensorboard graphs available during training... anything to look for there that indicates how the training is progressing?

Nnicearoni I've wondered about the avr_loss and tensorboard graphs available during trainin...

Furkan Gözükara SECourses•8/26/24, 10:42 PM

useless

Furkan Gözükara SECourses•8/26/24, 10:50 PM

multi gpu config and batch size statistics added : https://www.patreon.com/posts/110293257

Patreon

Kohya FLUX LoRA and Fine Tuning Training Full Tutorial For Local Wi...

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•8/26/24, 10:50 PM

you can literally get 1.4375 second / it with 4x gpu with minimal loss of quality

Furkan Gözükara SECourses•8/26/24, 10:51 PM

so 3000 steps takes around 72 minutes

Dazzastrous [4090]•8/26/24, 10:51 PM

fadounav•8/26/24, 10:54 PM

hahahaha xcellent how do you do that?

DDazzastrous [4090]Click to see attachment

fadounav•8/26/24, 10:54 PM

hahahaha xcellent how do you do that?

Ffadounav hahahaha xcellent how do you do that?

Furkan Gözükara SECourses•8/26/24, 10:56 PM

very likely trained my face

Dazzastrous [4090]•8/26/24, 10:58 PM

nah just prompted

in ideogram 2

Dazzastrous [4090]•8/26/24, 10:59 PM

its the best for text i find

�

𝔸𝖨ΞӾ•8/26/24, 11:01 PM

Midjourney is in shambles!

Furkan Gözükara SECourses•8/26/24, 11:01 PM

Horrid•8/26/24, 11:26 PM

seems the runpod .sh dosent work on regular linux ubuntu

Horrid•8/26/24, 11:26 PM

i tried so many times and it fails because it dosent recordnize what i want but its okay i may just rent a runpod server

HHorrid seems the runpod .sh dosent work on regular linux ubuntu

Furkan Gözükara SECourses•8/26/24, 11:27 PM

you need to have python c++ tools and cuda installed

Furkan Gözükara SECourses•8/26/24, 11:27 PM

runpod comes with pytorch docker template

Furkan Gözükara SECourses•8/26/24, 11:27 PM

if you can load it it would work

Horrid•8/26/24, 11:27 PM

i have cuda but not sure about c++

Furkan Gözükara SECourses•8/26/24, 11:27 PM

but it is hard to install those

Furkan Gözükara SECourses•8/26/24, 11:27 PM

also runpod doesnt have sudo command

Furkan Gözükara SECourses•8/26/24, 11:28 PM

you need to append them

Horrid•8/26/24, 11:29 PM

ill have to look into it tomorrrow

Darth Teethius•8/27/24, 12:38 AM

On the new flux lora test updates, i see speeds in s/it for each config json. Are those the speeds we shoukd be getting? Im running config 5 which is the one recommended for 12gb vram cards like my 3060. It was 12.2 s/it but im getting 25.5 s/it. Not sure what im doing wrong if those are the expected rates

DDarth Teethius On the new flux lora test updates, i see speeds in s/it for each config json. Ar...

Furkan Gözükara SECourses•8/27/24, 12:39 AM

no those are for A6000 GPU

Furkan Gözükara SECourses•8/27/24, 12:39 AM

relative speed

Furkan Gözükara SECourses•8/27/24, 12:40 AM

also on linux

Furkan Gözükara SECourses•8/27/24, 12:40 AM

25 second is looking normal

Furkan Gözükara SECourses•8/27/24, 12:40 AM

you have to reduce resolution for speed up but reduces quality

Furkan Gözükara SECourses•8/27/24, 12:40 AM

you can do 768x768

Darth Teethius•8/27/24, 12:42 AM

Thanks. Lol i thought i was way off. 25 seems fine

Maxivi•8/27/24, 12:45 AM

@Dr. Furkan Gözükara doesn't seem like comfy or forge accept loras trained with text encoder yet

MMaxivi @Dr. Furkan Gözükara doesn't seem like comfy or forge accept loras trained with ...

Furkan Gözükara SECourses•8/27/24, 12:47 AM

text encoder training didnt added to kohya yet

Furkan Gözükara SECourses•8/27/24, 12:47 AM

so i dont know :d

FFurkan Gözükara SECourses text encoder training didnt added to kohya yet

Maxivi•8/27/24, 12:49 AM

i got it to work

in sample images its fine, but they don't work on comfy or forge yet

Furkan Gözükara SECourses•8/27/24, 12:49 AM

well i am waiting kohya

Furkan Gözükara SECourses•8/27/24, 12:49 AM

he said he found today gonna add :d

Darth Teethius•8/27/24, 12:49 AM

For multi gpu training do the gpus need to be equal in vram? Like i have an 8 gb card im not using and a 12 gb card that is installed. Would the 8 gb card work for multi gpu training with the 12 gb one?

DDarth Teethius For multi gpu training do the gpus need to be equal in vram? Like i have an 8 gb...

Furkan Gözükara SECourses•8/27/24, 12:50 AM

very likely

Furkan Gözükara SECourses•8/27/24, 12:50 AM

it loads onto each gpu cloning basically

Darth Teethius•8/27/24, 12:50 AM

Looks like one trainer might be getting closer to flux release. Their dev branch is now called flux. So fingers crossed

Darth Teethius•8/27/24, 12:51 AM

Thanks

fadounav•8/27/24, 12:52 AM

anyone know how to add a lora on comfyUI for flux dev?

DDarth Teethius Looks like one trainer might be getting closer to flux release. Their dev branch...

Furkan Gözükara SECourses•8/27/24, 12:53 AM

yep

DDarth Teethius Looks like one trainer might be getting closer to flux release. Their dev branch...

Furkan Gözükara SECourses•8/27/24, 12:54 AM

our configs ready :d

fadounav•8/27/24, 12:54 AM

anyone know how to add a lora on comfyUI for flux dev? This is the workflow

Darth Teethius•8/27/24, 12:55 AM

You need to add a lora loader after the clip. Model from flux loader to left model port of lora loader. Clip out to lora loader and then model out to scheduler i believe

Darth Teethius•8/27/24, 12:55 AM

I have a basic workflow that i can post that shows it

Darth Teethius•8/27/24, 12:55 AM

Or post the image

DDarth Teethius You need to add a lora loader after the clip. Model from flux loader to left mod...

fadounav•8/27/24, 12:57 AM

please

Darth Teethius•8/27/24, 12:58 AM

https://civitai.com/models/618997/simpleadvanced-flux1-comfyui-workflows

[Simple+Advanced] FLUX.1 ComfyUI Workflows - LoRA Support v2 (No CL...

This workflow allows you to harness the power of FLUX.1 [dev] within ComfyUI. Follow these steps to set up and run the model. The true "simplest" v...

I've wondered about the avr_loss and tensorboard graphs available during training... anything to loo

Similar Threads

Similar Threads

Similar Threads