Software Engineering Courses (SECourses)•3y ago

So should I downgrade my Nvidia driver to 537 or leave it at 546.17 as it is? The new video card I

So should I downgrade my Nvidia driver to 537 or leave it at 546.17 as it is? The new video card I will be putting in is a 4070ti 12GB. I don't know if that's important or not.

Digital [Starburst]OP•11/25/23, 2:18 PM

I appear to have fixed the ControNet issue by reinstalling everything.

Digital [Starburst]OP•11/25/23, 2:19 PM

I need to test it however.

Aaesthetic hi, i just updated from 8gb to 16gb 4060ti. should i update my driver from 537 t...

Furkan Gözükara SECourses•11/25/23, 2:29 PM

yes i would update

Furkan Gözükara SECourses•11/25/23, 2:29 PM

with 16 GB you can do SDXL DreamBooth

DDigital [Starburst]So should I downgrade my Nvidia driver to 537 or leave it at 546.17 as it is? T...

Furkan Gözükara SECourses•11/25/23, 2:29 PM

use latest version

DDigital [Starburst]I appear to have fixed the ControNet issue by reinstalling everything.

Furkan Gözükara SECourses•11/25/23, 2:29 PM

great

Digital [Starburst]OP•11/25/23, 2:36 PM

Can you run Dreambooth with 12 GB?

Digital [Starburst]OP•11/25/23, 2:37 PM

Truthfully I didn't even know you could buy a 4060ti with 16 GB.

Digital [Starburst]OP•11/25/23, 2:38 PM

Thank you for answering my question about the driver.

Digital [Starburst]OP•11/25/23, 2:39 PM

One must not mess with SD settings when one does not know what one is doing.

Digital [Starburst]OP•11/25/23, 2:49 PM

Hmm...then Huggingface says, "Yes, it is possible to train DreamBooth on a 16 GB GPU using techniques like 8-bit Adam, fp16 training or gradient accumulation 1. However, fine-tuning the text encoder requires more memory, so a GPU with at least 24 GB of RAM is ideal.

Digital [Starburst]OP•11/25/23, 2:51 PM

So that would be a no. I cannot run Dreambooth with 12 GB of VRAM, or at least I can't expect anything better than pathetic performance.

Digital [Starburst]OP•11/25/23, 3:00 PM

Side point: For SDXL 1.0, the requirements are steeper. To support high-res 4K image generation, SDXL 1.0 requires GPUs with at least 12 GB of VRAM and a Windows machine.

DDigital [Starburst]Can you run Dreambooth with 12 GB?

Furkan Gözükara SECourses•11/25/23, 3:03 PM

yes you can run but only SD 1.5

DDigital [Starburst]Hmm...then Huggingface says, "Yes, it is possible to train DreamBooth on a 16 GB...

Furkan Gözükara SECourses•11/25/23, 3:03 PM

we are doing full SDXL DreamBooth with 15 GB VRAM on Kaggle

Digital [Starburst]OP•11/25/23, 3:19 PM

Okay

phillip•11/25/23, 3:24 PM

Can Stable Diffusion WebUI use a feature similar to kohya_ss's 'Print training command'?

Pphillip Can Stable Diffusion WebUI use a feature similar to kohya_ss's 'Print training c...

Furkan Gözükara SECourses•11/25/23, 3:24 PM

for what purpose?

FFurkan Gözükara SECourses for what purpose?

phillip•11/25/23, 3:25 PM

Because I am creating a batch shell script for stable diffusion.

phillip•11/25/23, 3:27 PM

I want to check the command for Stable Diffusion WebUI for my images and apply it to a shell script for automation.

Furkan Gözükara SECourses•11/25/23, 3:28 PM

i see . you can probably make but i don't know how to

rafraf•11/25/23, 3:47 PM

Hi @Dr. Furkan Gözükara is resize_images_v4 and cropper v7 still the most up to date scripts?

rafraf•11/25/23, 3:48 PM

the video is 5 months old

Rrafraf Hi @Dr. Furkan Gözükara is resize_images_v4 and cropper v7 still the most up to ...

Furkan Gözükara SECourses•11/25/23, 3:53 PM

they are very up to date

Furkan Gözükara SECourses•11/25/23, 3:53 PM

latest video here

Furkan Gözükara SECourses•11/25/23, 3:53 PM

https://youtu.be/Fbuyu35TkE4

YouTubeSECourses

SOTA Image PreProcessing Scripts For Stable Diffusion Training - Au...

One of the most important aspect of Stable Diffusion training is the preparation of training images. In this tutorial video I will show you how to fully automatically preprocess training images with perfect zoom, crop and resize. These scripts will hugely improve your training success and accuracy.

Scripts Download Link

https://www.patreon...

Furkan Gözükara SECourses•11/25/23, 3:53 PM

actually let me update 1 more time

Furkan Gözükara SECourses•11/25/23, 3:59 PM

updated to latest https://www.patreon.com/posts/88391247

Furkan Gözükara SECourses•11/25/23, 3:59 PM

@rafraf

rafraf•11/25/23, 4:00 PM

Ah yea I was following an older video on Patreon

rafraf•11/25/23, 4:02 PM

Btw @Dr. Furkan Gözükara do you know a windows software called One Commander? It’s an augmented file explorer

Rrafraf Btw @Dr. Furkan Gözükara do you know a windows software called One Commander? It...

Furkan Gözükara SECourses•11/25/23, 4:06 PM

sadly i dont know

Furkan Gözükara SECourses•11/25/23, 4:06 PM

also please redownload 25_Nov_2023_resizer.zip

Furkan Gözükara SECourses•11/25/23, 4:06 PM

or edit bat file and set cuda visible devices manually

740•11/25/23, 4:50 PM

hi guys! is animal lora training is the same as for humans? and what about regulation images?
I want to perpetuate the memory of my sick animal

FFurkan Gözükara SECourses also please redownload 25_Nov_2023_resizer.zip

rafraf•11/25/23, 5:16 PM

In the video you show these aspect ratios on white screen, but the official stable diffusion aspect ratios are shown in dark screen. I thought one needed not only to train in their aspect ratio but also their resolution?? @Dr. Furkan Gözükara

rafraf•11/25/23, 5:16 PM

1024 by 1280 is not in their list

rafraf•11/25/23, 5:16 PM

Does it mean, you consider only the aspect ratio important, not the resolution?

AIGambino•11/25/23, 5:28 PM

When having adetailer on, how can I make it so its only bringing detail of my guy and not everyone in a crowd photo

DR.Siva•11/25/23, 5:40 PM

kaggle error :

File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 47, in main
args.func(args)
File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/launch.py", line 977, in launch_command
multi_gpu_launcher(args)
File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/launch.py", line 646, in multi_gpu_launcher
distrib_run.run(args)
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/run.py", line 785, in run
elastic_launch(
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 134, in call
return launch_agent(self._config, self._entrypoint, list(args))
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 250, in launch_agent
raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
=====================================================
./sdxl_train.py FAILED
-----------------------------------------------------
Failures:
<NO_OTHER_FAILURES>
-----------------------------------------------------
Root Cause (first observed failure):
[0]:
time : 2023-11-25_17:18:32
host : 5d9e43f0dfc2
rank : 1 (local_rank: 1)
exitcode : -6 (pid: 1262)
error_file: <N/A>
traceback : Signal 6 (SIGABRT) received by PID 1262

Aldo Raine•11/25/23, 6:09 PM

I am getting the same thing using the patreon v12 file:

[E ProcessGroupNCCL.cpp:460] To avoid data inconsistency, we are taking the entire process down.
terminate called after throwing an instance of 'std::runtime_error'
what(): [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=3, OpType=ALLREDUCE, Timeout(ms)=1800000) ran for 1804412 milliseconds before timing out.
93%|████████████████████████████████████▎ | 1503/1616 [30:08<02:16, 1.21s/it]Traceback (most recent call last):
File "/opt/conda/bin/accelerate", line 8, in <module>
sys.exit(main())
File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 47, in main
args.func(args)
File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/launch.py", line 977, in launch_command
multi_gpu_launcher(args)
File "/opt/conda/lib/python3.10/site-packages/accelerate/commands/launch.py", line 646, in multi_gpu_launcher
distrib_run.run(args)
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/run.py", line 785, in run
elastic_launch(
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 134, in call
return launch_agent(self._config, self._entrypoint, list(args))
File "/opt/conda/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 250, in launch_agent
raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
=====================================================
./sdxl_train.py FAILED
-----------------------------------------------------
Failures:
<NO_OTHER_FAILURES>
-----------------------------------------------------
Root Cause (first observed failure):
[0]:
time : 2023-11-25_18:01:51
host : 10ed691951a5
rank : 1 (local_rank: 1)
exitcode : -6 (pid: 1077)
error_file: <N/A>
traceback : Signal 6 (SIGABRT) received by PID 1077
=====================================================