i say yes use xformers 0.0.17 and ema
i say yes use xformers 0.0.17 and ema
Fi havent checked. but you can use only man or photo of man as classification. it works even without classification but i found classification improves styling capability for faces
Fchange the loop in relauncher.py
Fcould be related to your training dataset
Fno backgrounds clothes should be same
Fmy computer also has 12 gb rtx 3060. it really looks good. you should be able to get good results same as base test on that video
VSo in youtube video you said the address we should run those commands ( uninstall ) is located in /workspace/venv/lib/python3.10
but for me there is no venv in workspace folder
I can find it in this address:
/workspace/stable-diffusion-webui/venv/lib/python3.10
but for me there is no venv in workspace folder
I can find it in this address:
/workspace/stable-diffusion-webui/venv/lib/python3.10
VIs that ok?
Fit works wherever you run actually
Fi mean on runpod at least
Fso it is okay
Fi think it is because of the template
Ftemplate links to venv so all command run on venv automatically
Vyes, I'm working on runpod. also I see some changes. For example I selected SD v1.5 template & after I opened it I see there is two models :
SDv1-5.ckpt
SDv2-768.ckpt
SDv1-5.ckpt
SDv2-768.ckpt
Fye
Fthey changed template
Fbut still works very well
Fcurrently i am doing 3 experiment
Fon dreambooth
Fon that template
Vand would you please answer to this question as well. your opinion is important to me 
I wanted to know have you tried the below format for learning rate field instead of 0.00001 ?
0.05:10,0.02:20,0.01:60,0.005:200,0.002:500,0.001:3000,0.0005
I wanted to know have you tried the below format for learning rate field instead of 0.00001 ?
0.05:10,0.02:20,0.01:60,0.005:200,0.002:500,0.001:3000,0.0005
Fi am using constant
Fthat can train very quickly
VI got this error 
VI run command the first for command without any issue
VI run the first four command without any issue *
Bi was actually talking about the content of the images and not the name of class
AI had my 40gb vram card run out of memory several times yesterday on Runpod. I switched to vast.ai and haven't faced an issue so far. However downloading from vast.ai is very slow.
Blike here, they arwe all real photos from unsplash
VThis command didn't work as well
Fthe images are generated by the name of class
Fit totally depends on configuration
Fif you want to fine tune model
Fimprove quality
Fthen you can feed better images
Fthe classification image logic is
Fharder to explain quickly
Fbut i am planning a video for that
VI used this command instead and it worked:
pip install xformers==0.0.17.dev464
Is it ok ?
pip install xformers==0.0.17.dev464
Is it ok ?
Bgotcha. also, one more thing. so I have the deleted the pod on which i trained yesterday. but now I want to fine tune the model even more. the steps will be the same?
Fyou need to upload same model
Fto continue training
For it will beging from 0
Boh yes that i'll upload, do i select this model as the base model in dreamboot and then create another new dreambooth model?
Fyep that is the idea
Bcool.
Bcan i use 2GPUs that'll give me 48gb vram and 83gb ram in total or i have to use one 1 gpu
FYou can't combine vram of them atm
FBut you can run 2 instance of web ui