Ok so I shouldn't add captions right?
Ok so I shouldn't add captions right?
Fwatch this section of this video
FYouTubeSECourses
Ultimate Kohya GUI FLUX LoRA training tutorial. This tutorial is product of non-stop 9 days research and training. I have trained over 73 FLUX LoRA models and analyzed all to prepare this tutorial video. The research still going on and hopefully the results will be significantly improved and latest configs and findings will be shared. Please wat...
F17:08 Dataset preparation, emphasizing image captioning, cropping, and resizing
F@m3dia_offline
Fyes i dont recommend
P@Furkan Gözükara SECourses i have 4 A6000 total 192 gb ram the training config will automatically distribute the load among all 4 right?
Pso the speed will be quiet fast for 48 gb 1 batch model?
Fyou need to use multi config
Fand only lora works
Fthis video shows logic and how to multi gpu https://youtu.be/-uhL2nW7Ddw
YouTubeSECourses
Unlock the power of FLUX LoRA training, even if you're short on GPUs or looking to boost speed and scale! This comprehensive guide takes you from novice to expert, showing you how to use Kohya GUI for creating top-notch FLUX LoRAs in the cloud. We'll cover everything: maximizing quality, optimizing speed, and finding the best deals. With our exc...
PReddit
Explore this post and more from the StableDiffusion community
PThey are claiming ultra photo realistic it would be really helpful
PFor generating photos of person
Fthese more looks like prompting related
Fbut let me test on myself new trained model an example
PYes please and let us know the results 
Fhow does this look didnt try prompt engineering just upscale of dreambooth model
F
Phi @Furkan Gözükara SECourses should I change the learning rate and max train steps for this configuration i am using 48gb batch 1 model and have 214 images ,let me know if i should change or is this perfect?
Fyou dont need to change anything
Fbut 300 epoch for 214 images can be overkill 
Fstill you are saving every 10 epoch so you can compare and find best checkpoint
�@Furkan Gözükara SECourses getting close to 5 minutes on 1 H100 full lora -
FNVME linked machine can use 48 GB VRAM amazing
amazing
Fon massed compute
P@Furkan Gözükara SECourses how much time will this take? or should I do 50 epochs and one checkpoint per 5 epochs?
Mdoes kohya supports traning for Flux Pro?
�flux pro is not open source and only available via API
Fit will take 114 hours
Fyes do 50 epochs
Fflux pro not published so not supporting
Fhere an important tips : https://www.reddit.com/r/SECourses/comments/1isdbw4/when_you_see_your_gpu_watt_usage_is_below_maximum/
Reddit
Explore this post and more from the SECourses community
F6 new experiments going on https://www.reddit.com/r/SECourses/comments/1iscy60/new_6_experiments_started_to_find_out_hopefully/
Reddit
Explore this post and more from the SECourses community
Fupdating my swarmui kaggle notebook it is working better than before
P@Furkan Gözükara SECourses for lora you told the formula to caluclate LR but for flux the LR is fixed to 0.000004 , i am doing 100 epichs and 10 checkpoints on H100
Pso should I keep the LR same 0.0000004
P?
P@Furkan Gözükara SECourses I am getting this error when I click on start training
P@Furkan Gözükara SECourses please reply I am stuck
Flooks like you didnt download or give model paths accurately
FLR changes only when you change batch size
PWhere to give model paths,I just downloaded models by the commands in the txt file and all the modesl are present in downloads folder
PAnd one more error koya isn't showing flux dev option after loading config
Fthat means you didnt do accurately
Fas i guessed
POk then should i delete this gpu and start everything fresh?
Fno you should download models accurately
Fand give their accurate path
Falso did you run kohya installer on massed compute?