this is typical of my 4090. He may have used xformers which I do not since I want higher quality
this is typical of my 4090. He may have used xformers which I do not since I want higher quality
VGotcha, thanks for the reply!
ALooking for the best workflow for a apparell mockup task:
I have a t-shirt design in transparent png and want to insert it onto a person wearing a plain t-shirt. The person is in a dynamic pose, so I need to get the shading, wrinkles done right.
Hope you can give me advice on how to approach this
I have a t-shirt design in transparent png and want to insert it onto a person wearing a plain t-shirt. The person is in a dynamic pose, so I need to get the shading, wrinkles done right.
Hope you can give me advice on how to approach this
SAnyone messed around with training TI embeddings? Trying to figure out why my TI + LoRA results are coming out with artifacts and over saturation. Seems like it could be a scheduler issue or VAE, but not sure right now.
VIs there a general rule for how well dreambooth training scales with multi-gpu? I used Dr Furkan's guide on a single 4090 and it took about 3 hours, decided to try it with a 6x 4090 setup, spawned it from cli with
... and the training only went down to 2 1/2 hours.
... and the training only went down to 2 1/2 hours.
Syou need to increase the batch size when doing multi gpu
Sie. increase the number of images processed at a time
VThanks, testing it now with a batch size of 6. I googled that earlier and the top comment I found said you want to keep a small batch size for multi gpu in order to make sure each GPU is always fully utilized, confusing
VHmm, so setting to anything higher than 2 leads to it crashing due to out of memory errors. However, with a batch size of 2 it went to a 4 hour estimate, whereas it was at 2.5 hours @ batch size of 1.
VHmm, could this be related? It seems like with 6 processes, batch size of 2 and multi-gpu enabled, I get 96 epochs and 100 batches per epoch. But with 1 process, batch size of 1 and multi-gpu disabled, I get 8 epochs @ 1200 batches per epoch.
VGPUs are being heavily utilized with 6 processes, batch size 2, multi-gpu. So it's not like they aren't doing anything. And yet, it's gonna take 4 hours and 15 minutes like this. Whereas with a single GPU it takes 3 hours.
V1.6s/it with 6 GPU, 1.07s/it with 1 GPU.
Fok time to check messages
Fhello welcome
Fcan you explain me more info
Fwhat you are trying to do
Fovertrained
MHello guys. I'm new to the discord and wanted to ask for information. I'm having a big problem these days, trying to train the lora sdxl on colab. Do you know who could give me a hand?
MThanks in advice 
F1.2 s / it is great
Fwelcome
Fsadly i have no colab for sdxl
Fbut we have automatic1111 web ui for kaggle
Fsupports all controlnet and sdxl
Fi did multi sdxl dreambooth and lora training on kaggle
Fhe is making some fixes to multi gpu recently
Mhi furkan! Thanks for answering me I also recently tried to use kaggle, but unfortunately I saw that they blocked it and it no longer works. Is there a way to bypass it? Thank you
I also recently tried to use kaggle, but unfortunately I saw that they blocked it and it no longer works. Is there a way to bypass it? Thank you
Lso how did you get it to actually offload across multiple gpus? I tried accelerate config, accelerate arguments etc
Fthere is bypass
Fon kaggle
Fit auto does that on kaggle
Fnow on runpod as well
FMSorry for asking but I've been going crazy for a week. Could you tell me how to solve it please?
Lwell I'm on local hardware haha, would've loved to know what runpod and kaggle do, that just scales across GPUs, though they might be just in SLI/nvlink
Fvector based search
Fi have script for this let me share on patreon
Bok amazing thankyou
Fare you trying on kaggle
For your pc?
Lmy own pc
Fwhat gpus you have?
L4090+3060
LLyou were part of that discussion, I tried to accelerate config the same way, but all that did was oom my 4090 even on batch 1 lol
Fhere
Fif you want exact duplicate
Fset that parameter
F