Hey lads, anyone know is its bad practices to train on images that are generated off the same one bu
Hey lads, anyone know is its bad practices to train on images that are generated off the same one but different (0.7) denoise.
Tfollowing
AIf you have low VRAM you basically have no choice but to use xformers, but if you have a decent amount of VRAM, you MUST disable it or it WILL negatively affect training quality.
TDoes anyone knows how to turn off rtx 3090 rgb lights. The light on GEFORCE RTX GameRock. Already have tried rgb fusion and geforece experience....
DHey all
Please i've been trying to use automatic 1111's dreambooth and i've been getting new errors i had never gotten before like this
Missing model directory, removing model: /workspace/stable-diffusion-webui/models/dreambooth/test1/working/unet
Please i've been trying to use automatic 1111's dreambooth and i've been getting new errors i had never gotten before like this
Missing model directory, removing model: /workspace/stable-diffusion-webui/models/dreambooth/test1/working/unet
Di'm also using it on runpod
ALooks like you did something wrong when configuring your model, try again by creating a new model and also click the "Unfreeze" button when you create your new model in Dreambooth.
Z@ashleyk I did the adan method posted in the article, i did 1600 steps, followed what he did, and my subject had no likeness to my dataset and not sure why
Zi used 17 images of my dcharacter
ADepends on the quality f your dataset. Dataset quality is EVERYTHING.
Z@ashleyk dataset looks great, i know it does cause i did the regular constant with warmup and did 8bitadam and it worked great, but then i tried adan dadaption with the dadap scheduler and nothing
Zi can post my json here if need be
Zdb_config.json3.99KB
AThat's incredible!
Looking forward to know more.
Looking forward to know more.
AWhy don't you just stick to 8bit Adam if that is working for you? Why do you specifically want to use the D-Adaptation optimizers? They are slower. I can't help you with LoRA training unfortunately, I haven't trained any LoRA yet. Maybe that Russian guy was just lucky with his specific dataset he was training using the Facebook D-Adaptation optimizers because I have not been successful in getting them to work yet with regular Dreambooth training either. I guess we need to wait for the good Dr. to test them out and make a video for us.
T@Dr. Furkan Gözükara how to resume training using last gpt.pth in DLAS cmd interface?
Zah ok gotcha
Z@ashleyk quick question about class instances per image, is that based on the amount of images i'm using, or the amount of steps i'm taking? What would be a good number of class instances per image if i'm using 50 images?
Zok so i think i have everything set right now, using 50 images for my dataset this time
db_config.json3.98KB
Zexcept for the xformers, i'm gonna change that to default
ANo clue on this, Anzhc told you yesterday, but I don't know myself.
Zyeah so dataset yesterday was 15, and it used my prior preservation images, today i used a dataset of 50, with 50 class instances per image, and wanted to generate 1750 images, as soon as I dropped that to 10, it said my number of class instances per image is sufficient
Zso the dadapt with lora only uses like 5 gb of vram, wow
Zthat's impressive
AWhich D-Adapt optimizer specifically? In my testing Adan D-Adapt used 32GB of VRAM for regular Dreambooth training vs less than 24GB for any other optimizer.
ZNot mine. I'm using adan dadapt
ZAnd it uses 5gb for lora extended
Z4.7 with xformers
T@Dr. Furkan Gözükara how to resume training using last gpt.pth in DLAS cmd interface?
AWhy would you enable xformers for training if you don't need it?
AGuess the either:
1) The extensions implementation of the D-Adaptation Optimizers for full Dreambooth training is not correct.
2) There is a bug in the D-Adaptation Optimizers on Linux.
1) The extensions implementation of the D-Adaptation Optimizers for full Dreambooth training is not correct.
2) There is a bug in the D-Adaptation Optimizers on Linux.
AThe Dreambooth extension has too many bugs, has too much code committed to dev branch that isn't even tested at all, and is just generally bad quality and has the worst SDLC of any project in the entire world so I have totally given up on it.
Zwell i mean it KINDA works, so i only used 15 images, and it was somewhat resembling my wife, but not 100%, so I upped the sample images to 51, and retesting
Zplus it took 4 hours to do 1600 steps yestereday and didn't resemble my wife at all, i've actually gotten better results with xformers,
Zi'm doing 5100 total steps this time, 50 epochs, 51 images dataset, will do an x/y/z and compare and see how it looks
Zwhat's weird is even thouhg i set lora unet leaerning rate to 1 according to the article, it trains at 1.10e-6, but also his article says that adan dapation ignores l/r so i'm assumign thats' why
AThe D-Adaptation optimizers adapt the Learning Rate constantly, hence their name. If you install Tensorboard, you will be able to view the graph of the LR and you will see that its not constant.
AHere is a Tensorboard graph of the Adan D-Adaptation LR, the LR becomes constant eventually but it basically has a bit of a warm up at the beginning.
Zyeah i did about 5100 steps andd still doesn't even remotely resemble my subject, so either it's broken, or you have to do an insane amount of steps
ZI may increase the epochs to like 150 and see what happens
Zhow do i pull up that graph on mine?
Fcan you eloborate more what you mean
Fhaha i would also turn off but never looked for
Fjust solved that problem
Fgit checkout main
Fin dreambooth extension folder
Fhopefully coming soon
Fin advanced tab there is a file path which is .pth. it is original file. we give our checkpoint there
Fin past we had huge problem with xformers
Fit was training but not learning
Fxformers 0.0.16