Software Engineering Courses (SECourses)•3y ago

Hey guys I'm a bit lost, can I use kohya to train model for SD 1.5? If yes, which one gives better r

Hey guys
I'm a bit lost, can I use kohya to train model for SD 1.5?
If yes, which one gives better result between Dreambooth and Kohya?

Elisoli•10/10/23, 5:43 PM

A6000 training sdxl kohya - 48gb config file, around 8000 steps, 3 hours makes sense?

EElisoli A6000 training sdxl kohya - 48gb config file, around 8000 steps, 3 hours makes s...

mtnmecca•10/10/23, 5:48 PM

yes

EElisoli If you put of a survey of most wanted tutorials i think itll maybe easier for yo...

Furkan Gözükara SECourses•10/10/23, 5:56 PM

good idea

Mmayna Hello, Started to follow your course on youtube for quality studio images, when...

Furkan Gözükara SECourses•10/10/23, 5:56 PM

i see there dreambooth and lora tabs

Furkan Gözükara SECourses•10/10/23, 5:56 PM

DDeleted User Running on local URL: http://127.0.0.1:7860 To create a public link, set `shar...

Furkan Gözükara SECourses•10/10/23, 5:56 PM

you need special notebook we shared on patreon

Furkan Gözükara SECourses•10/10/23, 5:57 PM

https://www.patreon.com/posts/run-on-free-like-88714330

Furkan Gözükara SECourses•10/10/23, 5:57 PM

https://www.patreon.com/posts/kohya-sdxl-lora-88397937

Vvirginia 1024 sdxl fail again ----------------------------------------------------------...

Furkan Gözükara SECourses•10/10/23, 5:57 PM

you really are not following what i am telling you

Aaelvelis Hi friends, I have this problem: Very slow image generation speed when using SDX...

Furkan Gözükara SECourses•10/10/23, 5:58 PM

add --medvram

Furkan Gözükara SECourses•10/10/23, 5:58 PM

assuming it is 12 gb vram

LLEGACIES Hey guys I'm a bit lost, can I use kohya to train model for SD 1.5? If yes, whic...

Furkan Gözükara SECourses•10/10/23, 5:58 PM

yes you can use. but dreambooth extension is better for sd 1.5

Furkan Gözükara SECourses•10/10/23, 5:59 PM

i will search best parameters for sd 1.5 too for kohya

Furkan Gözükara SECourses•10/10/23, 5:59 PM

after sdxl dreambooth hopefully

LEGACIESOP•10/10/23, 6:00 PM

do you have best setting for dreambooth model SD 1,5? I have take a rtx3090 on runpod since I only have 12gb vram on my pc

LLEGACIES do you have best setting for dreambooth model SD 1,5? I have take a rtx3090 on r...

Furkan Gözükara SECourses•10/10/23, 6:02 PM

i dont have yet sadly for kohya

Furkan Gözükara SECourses•10/10/23, 6:02 PM

it is on my research list

Furkan Gözükara SECourses•10/10/23, 6:02 PM

but i have for extension

Furkan Gözükara SECourses•10/10/23, 6:03 PM

https://www.patreon.com/posts/best-settings-of-85192985

Furkan Gözükara SECourses•10/10/23, 6:03 PM

https://www.patreon.com/posts/runpod-auto-84716845

LEGACIESOP•10/10/23, 6:03 PM

thanks was looking for dreambooth

Furkan Gözükara SECourses•10/10/23, 6:04 PM

1 auto installer

Furkan Gözükara SECourses•10/10/23, 6:04 PM

2 settings

LEGACIESOP•10/10/23, 6:04 PM

Is it compulsory to use class images?

LLEGACIES Is it compulsory to use class images?

Furkan Gözükara SECourses•10/10/23, 6:04 PM

not but improves

Furkan Gözükara SECourses•10/10/23, 6:04 PM

i have class images too

Furkan Gözükara SECourses•10/10/23, 6:04 PM

https://www.patreon.com/posts/massive-4k-woman-87700469

FFurkan Gözükara SECourses https://www.patreon.com/posts/best-settings-of-85192985

LEGACIESOP•10/10/23, 6:04 PM

why is there 2 links?

LLEGACIES why is there 2 links?

Furkan Gözükara SECourses•10/10/23, 6:05 PM

1 settings

Furkan Gözükara SECourses•10/10/23, 6:05 PM

1 auto installer

Furkan Gözükara SECourses•10/10/23, 6:05 PM

read all links

LEGACIESOP•10/10/23, 6:05 PM

ah ok thank you

LEGACIESOP•10/10/23, 6:05 PM

will go through it

FFurkan Gözükara SECourses assuming it is 12 gb vram

aelvelis•10/10/23, 6:13 PM

Yes, I ran the --medvram-sdxl command and it really helped for my 3060 ti. The only downside is the long wait for the first image to be generated. Also xformers helped me because opt-sdp-attention gives extremely low generation speed for sdxl.

Aaelvelis Yes, I ran the --medvram-sdxl command and it really helped for my 3060 ti. The o...

Furkan Gözükara SECourses•10/10/23, 6:17 PM

ah u have ti

Furkan Gözükara SECourses•10/10/23, 6:17 PM

it has only 6 gb vram

Narva•10/10/23, 6:18 PM

: ERROR Error while calling W&B API: user is not logged in (<Response [401]>)
Problem at: /workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py 294 init
wandb: ERROR The API key is either invalid or missing, or the host is incorrect. To resolve this issue, you may try running the 'wandb login --host [hostname]' command. The host defaults to 'https://api.wandb.ai' if not specified. (Error 401: Unauthorized)
Traceback (most recent call last):
File "/workspace/kohya_ss/./train_db.py", line 481, in <module>
train(args)
File "/workspace/kohya_ss/./train_db.py", line 258, in train
accelerator.init_trackers("dreambooth" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/accelerator.py", line 564, in _inner
return PartialState().on_main_process(function)(*args, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/accelerator.py", line 2096, in init_trackers
self.trackers.append(tracker_init(project_name, init_kwargs.get(str(tracker), {})))
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py", line 85, in execute_on_main_process
return function(self, *args, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/accelerate/tracking.py", line 294, in init
self.run = wandb.init(project=self.run_name, kwargs)
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1164, in init
raise e
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 1145, in init
run = wi.init()
File "/workspace/kohya_ss/venv/lib/python3.10/site-packages/wandb/sdk/wandb_init.py", line 764, in init
raise error
wandb.errors.AuthenticationError: The API key is either invalid or missing, or the host is incorrect. To resolve this issue, you may try running the 'wandb login --host [hostname]' command. The host defaults to 'https://api.wandb.ai' if not specified. (Error 401: Unauthorized)

Narva•10/10/23, 6:19 PM

I recently tried to start training in Kohya and I got that error, do you know what it could be?

Sandstormleader•10/10/23, 6:19 PM

is there anyway for me to run a seperate instance of automatic1111 so that i can have the old version with working dreambooth as well as a updated version @Dr. Furkan Gözükara