then make sure i click it inside of kohya?
then make sure i click it inside of kohya?
Fselect that model after loading config
Xokay but do i have to download it from hugging face an put it inside of my kohya file or i just need to click it inside of kohya because i see an option inside of kohya that i can click?
Fyou have to download it
Fand give its path
Ffollow tutorial video i recommend
Xok
XFi think not
Fthis came later
Fbut instead of flux dev 1.
Fyou pick this
Fsame
Xi will be sticking with flux dev 1 as config but should i click this or not any more
Fselect both
F
Fand use fp8 weight
Xgetting this error ERROR LoRA type must be set to 'Flux1', 'Flux1 OFT' or 'LyCORIS' if Flux1 checkbox is checked.
K@Dr. Furkan Gözükara doctor，please help me
Dim getting an error trying to use the supir upscaler that says it can't find the gradio module, but the installation said it was using the cached version on my system
KI have another error，it looks open input file failed，@Dr. Furkan Gözükara，and when I generated a video ，generate a new video the modle will reload ，it take a few time
Fyou didnt give correct model path
Fyou really should follow tutorial it shows
Fuse latest version please
Fby the way your video is generated perfectly fine
Fit is in outputs folder
Fjust your gradio did time out error
Fyou need to modify saved config file. those are defaults :d
Fif you make a fresh install and send me logs i can analyze
Fthere must have happened another error somewhere
SDr Furkan hi, i hope everything is ok, something change with kohya? some update?
Zhey @Dr. Furkan Gözükara whats the lastest video you have for training an SDXL model locally?
Zi see videos from a year ago - just wondering if there is anything newer
CSimilar to the most recent question.. I was wondering the same about local video generation. Everything in this space is moving so fast its hard to keep up with whats the best to use. Is anything better than Hailuo? Anything worth running off my own runpod?
Swan 2.1 looks like kling
Fkohya still same for flux
Fi dont have recent
Fnothing new i have . we have configs you can use them with flux installer kohya
Fthat works perfect
Fcurrently Wan 2.1 is very best
Fhopefully i will cover full training guide as well
Fsadly still no teacache today too :(/
Fnice torch getting updates for rtx 5000 series
Fi got to sleep now hopefully see you later
Kwhat course the gradio did time out error，should i fix something，the generated videos is great
TTo everyone using Wan 2.1 I'm just reminding that negative prompt is as important as regular prompt. It actually listens to negative prompt.
Take you time and craft a smart negative prompt. I have a preset one and tweak it on the fly when I notice things in a video I don't like.
Take you time and craft a smart negative prompt. I have a preset one and tweak it on the fly when I notice things in a video I don't like.
Fusually network connection or something in the windows system
Fnice. if you have a 1 for all negative prompt we can update our default
Fi am going to work on wan 2.1 training today
Jbut thats nothing regarding torchaudio ?
