but that doesn't work with constant learning rate right?
but that doesn't work with constant learning rate right?
ATraining sample images
ANo, the graph will be flat with constant LR.
FGitHub
Best Stable Diffusion and AI Tutorials, Guides, News, Tips and Tricks - Stable-Diffusion/Tutorials at main · FurkanGozukara/Stable-Diffusion
AWeights & Biases doesn't work with Dreambooth anyway. I logged a feature request for them to add it.
Fyes you can resume. use checkpoint pth in advanced options. you will see there
AThnak you
Ftortoise tts not related to stable diffusion
Fthank you so much. waiting youtube. i hope they restore :/
Fhello. i will make a new tutorial hopefully. with newest dreambooth extension torch 2 is necessary. meanwhile you can do this. update auto1111 to latest, updated dreambooth to latest, delete venv in workspace folder. restart auto1111
RPeople suck, they do this to channels so they get views. Yours was one of the 1st and honestly the best of all the tutorials.
Fi think that may be the issue. competitors reported
RI am having so much fun with roop. I'll have to subscribe to your patreon next pay. Your videos have been the best for SD etc.
ZYeah definitely would love a new video showing best settings for the new dreambooth optimizers @Dr. Furkan Gözükara
ZLora too if you can
Flook at this realism
Fnot even tested best checkpoint
F
RHello @Dr. Furkan Gözükara , which Ai model of SD V2 I need to use for training 768x768 pixel images ?
AKohya_ss dreambooth actually seems to train a person better for styling than the A1111 Dreambooth extension
AI couldn't get the example data set from YouTube trained properly with the Dreambooth extension, Kohya_ss is doing a much better job
Fsd 21 is really hard to train and get good results
Finteresting i think i should compare
Fby the way it turns out that 768x1024 loses styling ability significantly
Fgood for real images but not for style
Fand my training dataset had so many repeating cloth and background
Fthat degraded styling significantly
Fawesome ty so much
T@Dr. Furkan Gözükara in cloning, can I resume training Eg i left at 2550.gpt can I start again from that?
Twhy is this picking random voices? Can i use one specific tone only? Please guide
Fyes you can continue training from there
Frandom voices could be related to your training progress and dataset
TDoes more training means better result? which parameter to select to resume training?
TDataset contains only my voice of 16 mins and trained it for 5 hours until 2155_gpt.pth
F16 min voice could be low amount
Fusually more is better in training dataset as long as high quality
Awand.ai helps to determine where styling gets lost as well 
Fthe model learned even the whites in my beard haha
F
AThats awesome
Fbut there is too much memorization because of my training set
Fi should get good different bg and clothing pictures
AOh I thought you wanted it like that
Fit is good but cant stylze
TWhats the minimum amount of dataset? Also, will it resume if I start it like in video?
TI mean will dataset training resume?
Zthey need to hurry up aand make it so us lower vram guys can dreambooth
Zor should i just rent a runpod with someone with a beast card, train my models, then use them in auto
AYou don't need an 80GB beast though, 24GB is fine unless you want to use Adan D-Adapatation Optimizer, then you need at least 32GB