this is like a new stable difussion?
this is like a new stable difussion?
Fyes
Fdifferent model
Fcan follow prompts better
Dcan confirm when the estimated trime is beyond 30 min. I reduced the images to 13 , which made the caching latency time within 30min (23min to be exact) , it is working now
Depoch 1/2
steps: 20%|███ | 521/2600 [44:13<2:56:30, 5.09s/it, avr_loss=0.118]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00000521.safetensors
steps: 40%|████▊ | 1042/2600 [1:30:07<2:14:45, 5.19s/it, avr_loss=0.117]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00001042.safetensors
steps: 50%|██████ | 1300/2600 [1:53:08<1:53:08, 5.22s/it, avr_loss=0.113]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-000001.safetensors
epoch 2/2
steps: 60%|███████▏ | 1563/2600 [2:16:08<1:30:19, 5.23s/it, avr_loss=0.115]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00001563.safetensors
steps: 64%|███████▋ | 1677/2600 [2:27:02<1:20:55, 5.26s/it, avr_loss=0.114]
steps: 20%|███ | 521/2600 [44:13<2:56:30, 5.09s/it, avr_loss=0.118]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00000521.safetensors
steps: 40%|████▊ | 1042/2600 [1:30:07<2:14:45, 5.19s/it, avr_loss=0.117]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00001042.safetensors
steps: 50%|██████ | 1300/2600 [1:53:08<1:53:08, 5.22s/it, avr_loss=0.113]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-000001.safetensors
epoch 2/2
steps: 60%|███████▏ | 1563/2600 [2:16:08<1:30:19, 5.23s/it, avr_loss=0.115]
saving checkpoint: /kaggle/temp/models/DB_Gowthu-step00001563.safetensors
steps: 64%|███████▋ | 1677/2600 [2:27:02<1:20:55, 5.26s/it, avr_loss=0.114]
DSo i think there should be a disclaimer for now in your video to limit the images to 13-15 i guess. someone should work on this and see how many images can be within 30min. I am sure 18 images crosses 30 min
Fye maybe developer will fix it
Dluckily the notebook was on, i went into kohya gui and entered new data set(13 images) directory link and made changes to 521 steps and ran training
Dis it possible to move the generated tensor files to dataset directly ?
Die from temp files to dataset ?
Dnumbering of safetensor file is weird i guess ? 521/1042/1?/1563
Uhow is the speed of pixart-alpha?
DVaried , all types are important , but make it absolutely clear , in none of the photos other than face, nothing else repeats. Last time i trained my lora for my wife , i had few photos with same dress and it made a great impact on the outwhen when generating images. this time 
. waiting for dreambooth ...
. waiting for dreambooth ... 
Freally good
Fbut you need 16gb vram . i dont know if would work on 12 8 bit may work
Fi will try to make it work on kaggle as well
UNice. I think I'll try on runpod
Fyep should work great
UI saw this part on github page
Fye i checked it but
Ffor it to work
Ffor every image you load deload models
Fwill be ultra slow
Flike you lad text encoder, deload it , reload it everytime all models deleted and loaded again
UOh I see
Uthanks for your effort and great work
Fyou are welcome
Fi am testing runpod right now
Fa5000
Fupdated attachment again for runpod
Fneed 1 more update
F
B@Dr. Furkan Gözükara if I train for 1024x1024 resolution, can I also provide 2048x2048 or do you always recommend 1024x1024
FDid you have any success with multi_gpu training?
For me, In general, it seems to attempt launching training on multiple GPUs but then I get an out of memory error which terminates everything.
Training on 1 GPU with all the settings the same works. So it seems multiple GPU somewhere need more vmem on a single card
For me, In general, it seems to attempt launching training on multiple GPUs but then I get an out of memory error which terminates everything.
Training on 1 GPU with all the settings the same works. So it seems multiple GPU somewhere need more vmem on a single card
F1024x1024 i suggest
Fmaybe like 1024x1280
Fyes on kaggle i trained
Fit trains with dual gpu
SI will give it a try. to run it on runpod is with fast stable difussion? works with automatic 1111?
Fit is working with any since it makes own unique venv
Fi am trying to fix an error
Fon rupod
Fye we need to use
Fwith gradio share
Fon runpod
Swhat template should I pick on runpod? when you finish the repair?
Sthe same process as the interrogator?
Fany template . i selected same
Frunpod/pytorch
Fyes share working
