so what approach should I choose I want a specific pose for example that's hard to imitate with just
so what approach should I choose I want a specific pose for example that's hard to imitate with just generation?
Aalso @Furkan Gözükara SECourses is super realistic inpaint a thing in Flux?
.@Furkan Gözükara SECourses which dev branch of Kohya allows Flex1 training and do I have to change any parameters for it?
F1 : not sure how to do this flux
F2 : what you mean?
Fi replied. same as we trian flux
Fsd 3.flux1
.I've trained OpenFlux and re results are very good but it took much longer training than Flux dev dedistilled.
FRTX 5090 faster than L40S with FLUX at BF16
Fi am using kohya and it was even faster :d
Fbecause flex1 has lesser parameter
Falthough it probably need a better learning rate
Fwhich i dont plan to research atm
Fonly if a company or client requests
AI mean that is it possible to inpaint myself (Lora) on a picture I took with my phone and match perfectly?
Fi would say try and see
Fmay work
�Is that a full lora or dreambooth training?
Flora
Fkohya working on rtx 5090
Ftesting right now
�a full lora or single blocks?
Ffull
�Can you please make a video or a article on how to do this?
Fnope this is for clients. this is only useful for companies
Fi dont plan to share
�You used 2X H100 SXM?
�What if we use 1 H200 SXM?
Ahow much do you charge for it I'm interested
Fi am able to get same speed with it
Fi am doing research for 1000$ atm
�how much vram is it using?
Fi am using max in all cases
P@Furkan Gözükara SECourses hello ,I am trying to fine tune SDXL with LoRA with 23 high quality photos of a same person on civitai .Please let me know what parameter should i set.I am new to this so pleae let me know the values I should enter in each box.Thanks
Ftry to match with our kohya config
Pok
Rwhich is currently the best image to video model (free or paid, doesn't matter, just wanna know which is least bad!)
Fthe best ones are paid ones but i dont know which one best
Fkling really good
RI tried luma labs ray 2 and its not good
Rnearly every gen descends into some kinda terrible random motion nonsense, like it might do 2 seconds ok then suddenly someone goes flying through a wall or something stupid
Roccasionally it does something decent but a lot of times it just ignores what you ask and shit flies all over the place randomly, helicopters suddenly turn inside out and backwards lol
Rjust asking it to fly over a building following a helicopter lol
Fdoing new research https://www.reddit.com/r/SECourses/comments/1iqge85/started_doing_new_training_research_on_flux_dev/
Reddit
Explore this post and more from the SECourses community
DI'm trying out massed compute with your tutorial and the thindrives option says it's broken. any idea what that means and how to correct?
.@Furkan Gözükara SECourses I’m currently training Flex on my dataset and I feel it learns much quicker with half the steps compared to dev dedistilled. You were mentioning the learning rate might need adjustment. Do you think it needs to be lowered?
.After only 4000 steps. Not even finished the whole epoch.
A@Furkan Gözükara SECourses Hey so I experimented today a bit and got lost . Can you write what a segment syntax example looks like? Like is it 2 <segment> for -1 and -2?
What would the format look like ?
What would the format look like ?
AFor 2 Lora faces
Fif learning true fast yes