just used a different base model - Juggernaut Reborn instead of the the one mentioned on patreon
just used a different base model - Juggernaut Reborn instead of the the one mentioned on patreon
Fyes this is the reason
Fthese custom models are much harder to train
Fi think bigger LR or longer training might be necessary
Fi plan to investigate it
Fcan you train on SDXL 1.0 base?
Factually i will investitage this today
Fwhich custom model is best you think?
Mfor realism I think Juggernaut XL might be one of the best as well
Fok i will do my test on it
Ftime to rent a big pod :d
MThe creator seems to have a lot of support and has made a lot of progress
Mtraining wise I mean
Fyes support is necessary
Mso for 1.5 training, would you say it's a better idea to try training on the model you suggested on patreon?
Fyes that model is great
Fi tested 161 models
FYouTubeSECourses
In this video, I am excited to present a comprehensive tutorial featuring a script for downloading over 160 of the finest Stable Diffusion 1.5 (SD 1.5) models. In addition, I have meticulously tested and compared each of these 161 top-tier CivitAI SD 1.5 models, providing detailed commentary on their performance and capabilities.
Models Downloa...
Models Downloa...
MI've used to train in the past on Real Vision models, but I've been also using different tools. At the time Drembooth extension in A1111 worked fine then it was broken for a long time. Currently I've got Kohya and OneTrainer but each have some issues. OneTrainer doesn't offer LoRa export and in Khoya I cannot export Lora because of bits and bytes error that I never managed to fix xD
MI've done even a clean install of Kohya just in case but the export error remains
Fkohya working fine for me to export
Flet me test again
Falso kohya actively fixing errors
Fwe can report
MAre you using up to date version or some older, specific one?
Fusing latest right now
F22.6.2 testing
MOkay, I'll train once again, using the suggested model and see if that changes anything :)
MThank you for your answers and help!
Fye
F2024-03-01 00:11:15 WARNING A matching Triton is not available, some optimizations will not be enabled. init.py:55
Error caught was: DLL load failed while importing libtriton: The specified module could not be
found.
Error caught was: DLL load failed while importing libtriton: The specified module could not be
found.
Fgave error
Feasy to fix
Fbut still working
Fhaha
Moh that wasn't the error I was getting
Fit is just warning
MI had bit and bytes related error
Fyes working
Fi will try a fresh kohya install and see if still will work
Fi am extracting sdxl lora
F
F
MI can copy-paste the error later once I have a new model trained
FFhe stole your method too? xd
Flol the first comment
F
Fpeople is starting to hate you
Fas long as they dont use my scripts they can of course do