this is very useful : transformer-explainer : https://poloclub.github.io/transformer-explainer/

F

Furkan Gözükara SECoursesOP•8/14/24, 2:17 PM

multi line captions : https://github.com/kohya-ss/sd-scripts/blob/main/docs/config_README-en.md#multi-line-captions

GitHub

sd-scripts/docs/config_README-en.md at main · kohya-ss/sd-scripts

Contribute to kohya-ss/sd-scripts development by creating an account on GitHub.

F

Furkan Gözükara SECoursesOP•8/19/24, 12:49 PM

quantizations and can be applied to flux

F

Furkan Gözükara SECoursesOP•8/21/24, 8:50 PM

swarmui extra args --fast for 4xxx cards

F

Furkan Gözükara SECoursesOP•8/23/24, 12:50 AM

DeepSpeed / Deep Speed on Windows finally : https://github.com/microsoft/DeepSpeed/blob/master/blogs/windows/08-2024/README.md

GitHub

DeepSpeed/blogs/windows/08-2024/README.md at master · microsoft/Dee...

DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective. - microsoft/DeepSpeed

F

Furkan Gözükara SECoursesOP•8/30/24, 12:50 PM

swarmui inpaint face syntax

<segment:face-1,creativity,confidency threshold>

<segment:face-1,creativity,confidency threshold>

<segment:face-1,creativity,confidency threshold>

<segment:face-1,creativity,confidency threshold>

or in case of yolov8 models

<segment:yolo-filename.pt-1,creativity,confidency threshold>

<segment:yolo-filename.pt-1,creativity,confidency threshold>

<segment:yolo-filename.pt-1,creativity,confidency threshold>

<segment:yolo-filename.pt-1,creativity,confidency threshold>

F

Furkan Gözükara SECoursesOP•8/31/24, 1:53 PM

this is where you can continue training lora but i never tried to check if works or not

F

Furkan Gözükara SECoursesOP•8/31/24, 1:54 PM

F

Furkan Gözükara SECoursesOP•9/3/24, 1:44 PM

better face model for swarmui

https://huggingface.co/Bingsu/adetailer/resolve/main/face_yolov9c.pt

<segment:yolo-face_yolov9c.pt,0.7>

F

Furkan Gözükara SECoursesOP•9/3/24, 10:06 PM

Batch SUPIR Upscale Settings For FLUX

F

Furkan Gözükara SECoursesOP•9/5/24, 3:27 AM

accelerate extra args for multi multi gpu training : --main_process_port 30000

F

Furkan Gözükara SECoursesOP•9/10/24, 11:54 AM

GGUF (quantized) models on swarmui : https://github.com/mcmonkeyprojects/SwarmUI/blob/master/docs/Model%20Support.md#gguf-quantized-models

F

Furkan Gözükara SECoursesOP•9/11/24, 2:23 AM

you can set swarmui model files like this - put into diffusion_models folder

F

Furkan Gözükara SECoursesOP•9/11/24, 10:22 PM

'D:/Kohya-SS-GUI/kohya_ss' is on a file system that does not record ownership
To add an exception for this directory, call:

git config --global --add safe.directory D:/Kohya-SS-GUI/kohya_ss

F

Furkan Gözükara SECoursesOP•9/30/24, 10:58 PM

guys read this to learn a lot of important stuff about swarmui masking : https://github.com/mcmonkeyprojects/SwarmUI/issues/292

F

Furkan Gözükara SECoursesOP•10/9/24, 11:06 PM

https://www.gradio.app/playground

Gradio Playground

Play Around with Gradio Demos

F

Furkan Gözükara SECoursesOP•10/12/24, 11:33 AM

Currently, --highvram only affects caching of latens, and --lowvram only affects model loading. Training speed remains unchanged. - Kohya

F

Furkan Gözükara SECoursesOP•10/12/24, 10:47 PM

https://github.com/jakaline-dev/Triton_win/releases/tag/3.0.0

F

Furkan Gözükara SECoursesOP•10/14/24, 3:19 PM

F

Furkan Gözükara SECoursesOP•10/15/24, 8:06 PM

python -m xformers.info

python -m xformers.info

python -m xformers.info

python -m xformers.info

F

Furkan Gözükara SECoursesOP•10/15/24, 9:32 PM

screencapture-github-vladmandic-automatic-wiki-Models-2024-10-16-00_31_49.png

F

Furkan Gözükara SECoursesOP•10/17/24, 12:58 PM

https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#cuda-12x - onnx vs cudnn versions

F

Furkan Gözükara SECoursesOP•10/26/24, 11:04 AM

select cmd of kohya, hit ctrl +s to pause training, ctrl +q to continue

F

Furkan Gözükara SECoursesOP•10/27/24, 9:48 AM

To continue DreamBooth / Fine-Tuning, give your last checkpoint as base model

To continue LoRA training, give your last LoRA checkpoint here :

F

Furkan Gözükara SECoursesOP•10/29/24, 11:13 PM

man and woman different face classifier yolo models that can be used with swarmui : https://civitai.com/models/293448?modelVersionId=1007485
so you can give different prompt to man and woman faces automatically

F

Furkan Gözükara SECoursesOP•11/2/24, 7:42 PM

aria2c -x3 -s3 aria2c -x3 -s3 <URL> -o <FILENAME>

F

Furkan Gözükara SECoursesOP•11/3/24, 8:06 PM

GGFU vs FP8 speed

F

Furkan Gözükara SECoursesOP•11/6/24, 9:14 AM

excellent info regarding compile issues cuda paths : https://github.com/NVlabs/nvdiffrast/issues/46#issuecomment-2458715176

F

Furkan Gözükara SECoursesOP•11/13/24, 11:39 PM

very best non diffusion best upscaler : https://huggingface.co/Phips/4xRealWebPhoto_v4_dat2

Phips/4xRealWebPhoto_v4_dat2 · Hugging Face

F

Furkan Gözükara SECoursesOP•11/15/24, 1:39 PM

F

Furkan Gözükara SECoursesOP•11/16/24, 9:25 AM

F

Furkan Gözükara SECoursesOP•11/24/24, 12:42 AM

fastest thinlinc client settings

F

Furkan Gözükara SECoursesOP•11/24/24, 9:26 PM

regional prompt

F

Furkan Gözükara SECoursesOP•11/24/24, 9:26 PM

x and y define the starting point (top-left corner) of the region as percentages of the image dimensions (0 to 1).

x = 0 means the region starts from the far left.

y = 0 means the region starts from the top.

Values of x = 0.25 and y = 0.25 mean the region starts 25% from the left and top, respectively.

width and height define the size of the rectangular region, also as percentages of the total image dimensions.

A width = 0.5 and height = 0.5 mean the region spans 50% of the image’s total width and height.

x and y define the starting point (top-left corner) of the region as percentages of the image dimensions (0 to 1).

x = 0 means the region starts from the far left.

y = 0 means the region starts from the top.

Values of x = 0.25 and y = 0.25 mean the region starts 25% from the left and top, respectively.

width and height define the size of the rectangular region, also as percentages of the total image dimensions.

A width = 0.5 and height = 0.5 mean the region spans 50% of the image’s total width and height.

x and y define the starting point (top-left corner) of the region as percentages of the image dimensions (0 to 1).

x = 0 means the region starts from the far left.

y = 0 means the region starts from the top.

Values of x = 0.25 and y = 0.25 mean the region starts 25% from the left and top, respectively.

width and height define the size of the rectangular region, also as percentages of the total image dimensions.

A width = 0.5 and height = 0.5 mean the region spans 50% of the image’s total width and height.

x and y define the starting point (top-left corner) of the region as percentages of the image dimensions (0 to 1).

x = 0 means the region starts from the far left.

y = 0 means the region starts from the top.

Values of x = 0.25 and y = 0.25 mean the region starts 25% from the left and top, respectively.

width and height define the size of the rectangular region, also as percentages of the total image dimensions.

A width = 0.5 and height = 0.5 mean the region spans 50% of the image’s total width and height.

F

Furkan Gözükara SECoursesOP•11/25/24, 11:28 PM

FLUX.1 [dev] GPU performance comparison

F

Furkan Gözükara SECoursesOP•11/25/24, 11:29 PM

flux-1-dev-gpu-performance-comparison-v0-nmbja6wvo43e1.png

F

Furkan Gözükara SECoursesOP•11/28/24, 10:18 AM

latent upscale with flux - recommended percent 0.5-0.6

F

Furkan Gözükara SECoursesOP•11/29/24, 12:42 AM

who cant purchase patreon i will accept crypto and gift you

https://www.patreon.com/SECourses/gift

F

Furkan Gözükara SECoursesOP•12/5/24, 12:02 PM

who gets lower quality on SwarmUI uncheck this box

F

Furkan Gözükara SECoursesOP•12/5/24, 12:46 PM

New UI feature: confused about the more complex prompt syntax inputs like Segment or Regional prompts? Be confused no more, there's now helper UIs to guide you through constructing perfect inputs every time!

F

Furkan Gözükara SECoursesOP•12/6/24, 11:59 PM

flash attention wheels : https://github.com/bdashore3/flash-attention/releases

F

Furkan Gözükara SECoursesOP•12/18/24, 11:18 PM

https://github.com/facebookresearch/pytorch3d/discussions/1752#discussioncomment-11152790

F

Furkan Gözükara SECoursesOP•1/8/25, 12:04 PM

F

Furkan Gözükara SECoursesOP•1/8/25, 10:47 PM

how to load config on runpod

F

Furkan Gözükara SECoursesOP•1/12/25, 12:56 AM

how to generate pre-compiled wheel

pip install wheel
pip wheel --no-deps -w dist .
inside repo folder
it will be saved on dist folder

pip install wheel
pip wheel --no-deps -w dist .
inside repo folder
it will be saved on dist folder

pip install wheel
pip wheel --no-deps -w dist .
inside repo folder
it will be saved on dist folder

pip install wheel
pip wheel --no-deps -w dist .
inside repo folder
it will be saved on dist folder

this one worked on xpose
pip install setuptools wheel
python setup.py sdist bdist_wheel

this one worked on xpose
pip install setuptools wheel
python setup.py sdist bdist_wheel

this one worked on xpose
pip install setuptools wheel
python setup.py sdist bdist_wheel

this one worked on xpose
pip install setuptools wheel
python setup.py sdist bdist_wheel

F

Furkan Gözükara SECoursesOP•1/26/25, 2:04 AM

flux image to image upscale

F

Furkan Gözükara SECoursesOP•1/29/25, 12:21 AM

F

Furkan Gözükara SECoursesOP•2/14/25, 7:54 PM

flux fill inpainting model training : https://github.com/nftblackmagic/catvton-flux

F

Furkan Gözükara SECoursesOP•2/20/25, 12:02 AM

for sage attention, use --use-sage-attention in the Backends tab Extra Params

F

Furkan Gözükara SECoursesOP•2/22/25, 11:19 AM

nightly torch insall

python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

python -m pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

F

Furkan Gözükara SECoursesOP•2/22/25, 4:35 PM

example video generation model prompt test

Stop-motion, model animation. A time-machine on the table whirs into life with pistons moving and gears spinning. The hands of the clock spin rapidly. The characters watch the machine in awe. The camera orbits rapidly around the machine, showing the professor's Victorian style laboratory from all angles. The machine sparks and buzzes with waves of visible electrical energy.

Stop-motion, model animation. A time-machine on the table whirs into life with pistons moving and gears spinning. The hands of the clock spin rapidly. The characters watch the machine in awe. The camera orbits rapidly around the machine, showing the professor's Victorian style laboratory from all angles. The machine sparks and buzzes with waves of visible electrical energy.

Stop-motion, model animation. A time-machine on the table whirs into life with pistons moving and gears spinning. The hands of the clock spin rapidly. The characters watch the machine in awe. The camera orbits rapidly around the machine, showing the professor's Victorian style laboratory from all angles. The machine sparks and buzzes with waves of visible electrical energy.

Stop-motion, model animation. A time-machine on the table whirs into life with pistons moving and gears spinning. The hands of the clock spin rapidly. The characters watch the machine in awe. The camera orbits rapidly around the machine, showing the professor's Victorian style laboratory from all angles. The machine sparks and buzzes with waves of visible electrical energy.

this is very useful : transformer-explainer : https://poloclub.github.io/transformer-explainer/

Similar Threads

Similar Threads

Similar Threads