I haven’t yet look the workflow for SDXL yet, is it possible to train dreambooth model using a 4090
I haven’t yet look the workflow for SDXL yet, is it possible to train dreambooth model using a 4090? Or do I have to rent a runpod for that?
Jw/ a 4090 you can train locally using DB - basically need 24gb.
AHello @Dr. Furkan Gözükara. I have used s3fs to mount aws s3 bucket onto system. I have used kohya_ss for training using the images from s3fs mounted folder which was successfull. I've mentioned the s3fs folder in additional networks path in settings of sd-webui, but lora files are not showing although they exist in folder.
MHey @Dr. Furkan Gözükara , do you have a .JSON / .TOML file for the best parameters to use in KOhya SS to train a very realistic Model ( i prefere Photon / Realistic Vision V5 )
The file i'm talking about is the one containing all these parameters .
( Alpha , fp16 .; prior_loss_weight = resolution = enable_bucket = min_bucket_reso = max_bucket_reso = .. )
is it free or is it only for patreon guys ??
The file i'm talking about is the one containing all these parameters .
( Alpha , fp16 .; prior_loss_weight = resolution = enable_bucket = min_bucket_reso = max_bucket_reso = .. )
is it free or is it only for patreon guys ??
Enot good either
Ecfg 5 feels like better results but not so flexiable, might have over trained it from the start somehow?
Ealso Adetailer gives corrupted faces
Sgit config --global --add safe.directory I:/auto1111/stable-diffusion-webui
Creating venv in directory I:\auto1111\stable-diffusion-webui\venv using python "C:\Users\The_A\AppData\Local\Programs\Python\Python310\python.exe"
venv "I:\auto1111\stable-diffusion-webui\venv\Scripts\Python.exe"
fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui'
'I:/auto1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:
git config --global --add safe.directory I:/auto1111/stable-diffusion-webui
fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui'
'I:/auto1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:
Creating venv in directory I:\auto1111\stable-diffusion-webui\venv using python "C:\Users\The_A\AppData\Local\Programs\Python\Python310\python.exe"
venv "I:\auto1111\stable-diffusion-webui\venv\Scripts\Python.exe"
fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui'
'I:/auto1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:
git config --global --add safe.directory I:/auto1111/stable-diffusion-webui
fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui'
'I:/auto1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:
Sthis is what i get when i try to use the auto installer on enxternal hdd @Dr. Furkan Gözükara
SFile "I:\auto1111\stable-diffusion-webui\modules\launch_utils.py", line 192, in git_clone
run(f'"{git}" -C "{dir}" checkout {commithash}', None, "Couldn't checkout {name}'s hash: {commithash}")
File "I:\auto1111\stable-diffusion-webui\modules\launch_utils.py", line 115, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't checkout {name}'s hash: {commithash}.
Command: "git" -C "I:\auto1111\stable-diffusion-webui\repositories\stable-diffusion-stability-ai" checkout cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf
Error code: 128
stderr: fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai'
'I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai' is on a file system that does not record ownership
To add an exception for this directory, call:
git config --global --add safe.directory I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai
Press any key to continue . . .
run(f'"{git}" -C "{dir}" checkout {commithash}', None, "Couldn't checkout {name}'s hash: {commithash}")
File "I:\auto1111\stable-diffusion-webui\modules\launch_utils.py", line 115, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't checkout {name}'s hash: {commithash}.
Command: "git" -C "I:\auto1111\stable-diffusion-webui\repositories\stable-diffusion-stability-ai" checkout cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf
Error code: 128
stderr: fatal: detected dubious ownership in repository at 'I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai'
'I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai' is on a file system that does not record ownership
To add an exception for this directory, call:
git config --global --add safe.directory I:/auto1111/stable-diffusion-webui/repositories/stable-diffusion-stability-ai
Press any key to continue . . .
Fof course
F4090 3090 both are amazing
Feven with 16 gb when xformers enabled
Fi dont know
Fmust be path error or script error maybe
Fyes i have on patreon
Fyou can use
F
Fi dont see noise. i think related to training dataset
Fpermission error
Fprobably you need to change permissions of your external hdd
S@Dr. Furkan Gözükara any tips for that?
Fi tested my auto installer on external
Fit worked
Fyes from advanced security tab
Fyou can change
F
E8bit 10gb vram bat file using blip2 opt6.7b coco
Fwhat is your GPU?
E3060 RTX 12gb
Fhow much vram usage you have before starting it?
Fit must be using shared ram
Eyes could be , after effects from adobe was running the same time i will try again with cleaner ram
btw any difference between blip2-6.7 and coco version ?
btw any difference between blip2-6.7 and coco version ?
Fnot much
Fi am still testing
Mjust became a new patreon 
Mthis is only for SD XL ?
Fthank you so much
Ffor sd 1.5 i suggest dreambooth extension
Fhere the auto installer link and settings
Fdreambooth extension supporting EMA
FKohya will also hopefully implement it
FI will also research Kohya best settings for SD 1.5 too
MSo the .json file is within the autoinstaller ? it comes with Parameters built in ? since it's in auto ?
I don't see the JSon file if not
I don't see the JSon file if not
Ebetter now clean ram
MI just trained a Lora model on RealisticVision with 32/32 (NR/NA) and Prodigy with these parameters:
decouple=True use_bias_correction=True safeguard_warmup=True
It's amazing to achieve professional results with such low weights on a SD 1.5 model. Here is a sample (original and hires fixed):
decouple=True use_bias_correction=True safeguard_warmup=True
It's amazing to achieve professional results with such low weights on a SD 1.5 model. Here is a sample (original and hires fixed):
Mi do prefer lora Kohya ss and I also started following your video ( produce studio quality images)
I installed the Kohysa ss Gui , now I only need the setup parameters ( if you have any json file ready )
is the one in this link work with kohya ss ?
I installed the Kohysa ss Gui , now I only need the setup parameters ( if you have any json file ready )
is the one in this link work with kohya ss ?
