q6 might be better
q6 might be better
Fi recommend test and compare both
Fq8 really high quality almost bf16
Zgoing to have to wait for a tutorial - img2video is still a bit confusing to me
Fye a little bit harder i am also testing :d
Zill wait for your results!
Zwhen downloading the model - does it download everything we need to use it?
Fthis downloader + some other things swarmui auto handles
Fat first generation
Fso you dont manually download anything
Zsounds good. ill give the Q6 model a try
FPatreon
Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon
Fi will update skyreels config images when i get them best
Zgetting an error with GGUF support installed when trying to run
Fyes you need to install a node
Fit asks when you first time ruın
Finstall it
Fclick load model it asks
Zi didnt see anything but ill try with the new launcher. do we just drop and drop all the files into our old folder?
Fuse downloader to download model
Fwait i will show
Fhugging face such a beast
Zi got the model - i turned on img2video - but how do we get it to actually generate?
Fwait tutorial
Zhmm, i never got this message
Zis there a different button to run the img2video - or do we just hit the normal generate button?
F
Zill try to download a different model and reboot swarmUI and see if i get the message
Z
1how is Hunyuan running on the 5090?
Fgood
Fwill show in video
1Perfect timing, just finishing rig upgrade stm
Fnice
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dnice but I dont think you have enough light in there 
Zthis PC isnt getting hot anymore soon!
GHi @Dr. Furkan Gözükara, last year I trained a bunch of SDXL dreambooth models using your Kaggle notebook here: https://www.patreon.com/posts/kohya-sd-1-5-88397937. I'm just getting back into it, but having trouble. Firstly, after cloning https://github.com/FurkanGozukara/kohya_ss, the script sdxl_train.py doesn't seem to exist. I fixed this by updating '!git clone ...' to '!git clone --recursive ...' in the first cell, but the script /kaggle/working/kohya_ss/sd-scripts/sdxl_train.py now crashes with the error message 'RuntimeError: operator torchvision::nms does not exist'. Any ideas on how to fix this? (Patreon user 'Michael Dann' fyi.)
Fare you using v22?
GYup
Fare you sure you are using v22 kaggle notebook?
Flet me update the notebook to v23
GYeah I'm sure:
Glegend
Ftrying to fix
Fthere is an annoying gradio error 
Fye not working atm i opened 2 issues
Fi got to sleep
Gno problem at all, appreciate you taking a look at it
GI managed to hack the notebook and get it working by avoiding gradio altogether (just manually creating the input folder structure) so no longer a pressing issue