Software Engineering Courses (SECourses)•3y ago

by the way spot is not dependable to use for training

FFurkan Gözükara SECourses by the way spot is not dependable to use for training

ashley•6/5/23, 11:44 PM

Yes they can take it back at any time but I have been using it the whole time, been lucky I guess

felixsanz•6/5/23, 11:44 PM

i never use spot because i don't want to get disconnected

Ffelixsanz i never use spot because i don't want to get disconnected

Furkan Gözükara SECoursesOP•6/5/23, 11:44 PM

i tried once i think :d

felixsanz•6/5/23, 11:44 PM

not sure if it's happening or not

Furkan Gözükara SECoursesOP•6/5/23, 11:45 PM

i love raid 1

Furkan Gözükara SECoursesOP•6/5/23, 11:45 PM

one of my hdd in my game server had become like this

ashley•6/5/23, 11:45 PM

My current spot pod has been up for almost 5 hours

Furkan Gözükara SECoursesOP•6/5/23, 11:45 PM

Furkan Gözükara SECoursesOP•6/5/23, 11:45 PM

they replaced it, raid 1 auto did rebuild

Furkan Gözükara SECoursesOP•6/5/23, 11:46 PM

the other disk is also failing

Furkan Gözükara SECoursesOP•6/5/23, 11:46 PM

they will replace it too

Furkan Gözükara SECoursesOP•6/5/23, 11:46 PM

Furkan Gözükara SECoursesOP•6/5/23, 11:47 PM

due to 8% hdd it was reading and writing like with 5mb per second

Furkan Gözükara SECoursesOP•6/5/23, 11:47 PM

now back to 500-1000

Pplayereckksdee @Dr. Furkan GözükaraI have noticed that the quality of training on kohya is wors...

playereckksdee•6/5/23, 11:52 PM

It might have been training incorrectly but still giving better results, I think current runpod kohya actually trains correctly but something with runpod changed and fixed the issue with training and made it train as it should

ashley•6/5/23, 11:56 PM

Looks like Network Rank was removed from Kohya_ss

Pplayereckksdee It might have been training incorrectly but still giving better results, I think...

ashley•6/5/23, 11:57 PM

That would have been an issue with Kohya code, not RunPod

Aashley That would have been an issue with Kohya code, not RunPod

playereckksdee•6/5/23, 11:58 PM

Previous versions of kohya would also give same result

Nom•6/6/23, 1:31 AM

@Dr. Furkan Gözükara what are your thoughts on train styling settings? Is it still the same settings as your youtube video? Lion optimizer with step ratio 0.5 and decay of 0.1?

NNom @Dr. Furkan Gözükara what are your thoughts on train styling settings? Is it sti...

Furkan Gözükara SECoursesOP•6/6/23, 1:33 AM

until new methods tested and verified to be better yes

Furkan Gözükara SECoursesOP•6/6/23, 1:34 AM

started training myself with my newest workflow

Furkan Gözükara SECoursesOP•6/6/23, 1:34 AM

Furkan Gözükara SECoursesOP•6/6/23, 1:34 AM

using 18.3 gb vram

Furkan Gözükara SECoursesOP•6/6/23, 1:34 AM

not using xformers

Furkan Gözükara SECoursesOP•6/6/23, 1:34 AM

going to sleep hopefully see you tomorrow

FFurkan Gözükara SECourses Click to see attachment

ashley•6/6/23, 1:41 AM

I miss the training summary before training starts, I wish they would bring that back and not make it debug log level

ashley•6/6/23, 1:47 AM

Anyone know why I am suddenly getting this error in Kohya_ss? It was working fine earlier:

    self.tk = _tkinter.create(screenName, baseName, className, interactive, wantobjects, useTk, sync, use)
_tkinter.TclError: no display name and no $DISPLAY environment variable

    self.tk = _tkinter.create(screenName, baseName, className, interactive, wantobjects, useTk, sync, use)
_tkinter.TclError: no display name and no $DISPLAY environment variable

FFurkan Gözükara SECourses shared in my video description

ashley•6/6/23, 2:23 AM

Which video? Can't find it in this video: https://www.youtube.com/watch?v=3uzCNrQao3o

YouTubeSECourses

How To Install And Use Kohya LoRA GUI / Web UI on RunPod IO With St...

How to install famous Kohya SS LoRA GUI on RunPod IO pods and do training on cloud seamlessly as in your PC. Then use Automatic1111 Web UI to generate images with your trained LoRA files. Everything is explained step by step and amazing resource GitHub file is provided with necessary commands. If you want to use Kohya's Stable Diffusion trainers...

teachtaha•6/6/23, 6:21 AM

@Dr. Furkan Gözükara Will it work if I replace tortoise fast with just tortoise tts in stable diffusion? for better quality..

teachtaha•6/6/23, 6:34 AM

@Dr. Furkan Gözükara whenever I start training, it starts from 0 epoch, is there any way I can resume training. like a neural network in which you train the machine and it gets better day by day

Kasmos•6/6/23, 7:34 AM

Hi kind of all Does anyone know if this is a hardware issue?

Furkan Gözükara SECoursesOP•6/6/23, 10:00 AM

dmn youtube removed my video
the message was

"Showing viewers how to use apps, websites, or other information technology to gain unauthorized access to audio content, audiovisual content, full video games, software, or streaming services that normally require payment."
appealed i hope they approve my appeal
i wonder it was result of people reported or automated or manual

ashley•6/6/23, 10:06 AM

Probably people reporting it. People find reasons to be offended about anythintg these days, as though there aren't enough REAL problems in the world.

ashley•6/6/23, 10:07 AM

I once posted a joke on Facebook and Facebook suspended my account because someone who didn't have a sense of humour reported it, so I no longer have Facebook, and I am banned from Twitter as well

Furkan Gözükara SECoursesOP•6/6/23, 10:10 AM

i appealed i hope they grant it

FFurkan Gözükara SECourses i appealed i hope they grant it

ashley•6/6/23, 10:12 AM

Yeah, it didn't seem to do any of those things they said, unless they consider deep faking somehow providing unauthorized access to content.

FFurkan Gözükara SECourses i appealed i hope they grant it

Redman•6/6/23, 10:51 AM

Man, I just went to rewatch the part about audio sync issues and running into an issue. That sucks your tutorials are the best

unlimitless•6/6/23, 10:56 AM

hi boss, after the upgraded webui on runpod, i did re-downgrade torch to 1.13.1 and all other things as you told us on your video, and every thing goes well. but after i finished my lora training on dreambooth and tring to test it on webui, this hapened : "RuntimeError: Couldn't load custom C++ ops. This can happen if your PyTorch and torchvision versions are incompatible, or if you had errors while compiling torchvision from source. For further information on the compatible versions, check https://github.com/pytorch/vision#installation for the compatibility matrix. Please check your PyTorch version with torch.version and your torchvision version with torchvision.version and verify if they are compatible, and if not please reinstall torchvision so that it matches your PyTorch install." i tried to upgrade torch to the newest vision but the system tell me the vision already exist. what should i do

GitHub

GitHub - pytorch/vision: Datasets, Transforms and Models specific t...

Datasets, Transforms and Models specific to Computer Vision - GitHub - pytorch/vision: Datasets, Transforms and Models specific to Computer Vision

ashley•6/6/23, 10:57 AM

Weights & Biases (https://wandb.ai/) is pretty cool for tracking your training experiments with things like Kohya_ss and EveryDream2trainer. I don't think Dreambooth supports it yet.

Weights & Biases – Developer tools for ML

WandB is a central dashboard to keep track of your hyperparameters, system metrics, and predictions so you can compare models live, and share your findings.

unlimitless•6/6/23, 11:00 AM

Uunlimitless Click to see attachment

ashley•6/6/23, 11:06 AM

Answered you on RunPod Discord.

RRedman Man, I just went to rewatch the part about audio sync issues and running into an...

Redman•6/6/23, 11:10 AM

Side note you can write click the imported frames, click on clip attributes and edit to original video fps.... Might help someone at some point. Or is that what the answer was in the tutorial lol