know new problem when i start to launch my "train" i got this
know new problem when i start to launch my "train" i got this
message.txt5.3KB
JOutOfMemoryError: CUDA
Jnot enough vram
J"GPU 0; 8.00 GiB total capacity; 5.50 GiB already allocated; 0 bytes free; 7.12 GiB reserved in total by PyTorch"
Cmmmh fixable ?
Fout of memory error
Jthe zero to hero video mentions a wiki page by dreamboth that lists vram reduction options.
Fyou can do only lora not dreambooth
Coh so sad i was think understand that with xform and adam 8 GVRAM
F8 gb for lora
Fyou need 12 for dreambooth
Fbase model
Fvs my style teaching
Cok thanks for the information
JYou could buy calculating time or use the googel service as a test version for free.
Cys i do that before
Fexact same settings base model vs my mj level style
Fuploading model to post on patreon
Cwith google collab
Clora need other ajustment
JIs it typical that it looks slightly foggy with this style or is that just a random thing in this example?
JThe repair was unsuccessful.
I think I'll strate with a fresh version. To make every mistake again
I think I'll strate with a fresh version. To make every mistake again
Fi think random
Fusually best solution :d
JDoes torch 2 only improve performance or also quality?
If only performanc, then I would leave it out for now. I have enough power that that wouldn't be too bad.
If only performanc, then I would leave it out for now. I have enough power that that wouldn't be too bad.
Fonly performance atm i believe
Fit is still in development
Jthanks
Fmodel file posted on patreon for anyone interested in
Zwhat are the two packages we need to upgrade in stable diffusion for faster processing? can we not just pip install --upgrade then the package names?
Ztransformers and can't remember the other one
Fwell you need pre compiled wheel
Fso you cant just execute install command not yet
Zis there any prompt to activate face restoration?
FHi! I´m learning how to train models following this tutorials. I´m training my first one and in the preview output there are 4 images the first one is super random, nightmarish, the second one looks like the person I´m trying to model and then there are two graphics that I don´t understand and don´t know how to real:
F
FIs this common?
SHi I am having all sorts of problems so i decided to try and make an embed because it seemed easier but they are all coming out like they were painted by a madman. Is this normal?
FI read somewhere that the latest Dreambooth is just broken, no matter what you do, what parameters you adjust the results would be nightmarish
FIt´s so frustrating! Spent days trying for a model to work and to understand the app but the results are always garbage after hours of waiting
FWhat´s an embed?
SAn embed is a way of training that, according to my research, allows you to make a model based on one of the models (say, 1.5) and then you can use that face in any model that is based on 1.5.
So it would allow one to put the character they trained into a lot of things. Unless they look like claymation horrors from hell's basement.
So it would allow one to put the character they trained into a lot of things. Unless they look like claymation horrors from hell's basement.
FHo yes! That´s what I´m doing!
FI´m starting to get interesting results FINALLY!!!
Sdoes "interesting" mean melted abomination? Cause if so, me too!
SThe Doc has a video on training a LORA but he never mentions an embed so maybe it's trash?
FI think that "Save your Model Frequency" Is a very important value, I did just that, the memory usage goes crazy because at the end of the training you have lots of models. I assumed as a noob that the more the code works on a model the better it gets but that´s not the case, I think its called overtraining so... I ended with 10 versions of a model and apparently the third one is working very good, the fourth one gives bad results the fifth one very bad results and so on...
FNo! realistic results!
Spfft, if that's what you're going for, I suppose. But it's blurry nonsense for me! LET'S GO!
FHo! no, I he mentions that LORA is more difficult so I´m starting with embed 1.5
F