another question. does fp8 makes it faster?
another question. does fp8 makes it faster?
Cokk
Fnope only reduces RAM and VRAM
Fbecause operations are still made in bf16
Fonly model weights loaded as fp8
Ci see.
Cthanks !
AI made a mistake and used the 720 image to video 
1 hour later i'm only 65% done
1 hour later i'm only 65% doneADoes having a large base image resolution or the output resolution being different than the incoming image resolution make a difference in terms of speed for Wan i2v?
Zmessage.txt2.42KB
Znevermind models havent' downloaded lol
Awhere would i clone qwen to if I want to use prompt enhance? For some reason even if I remove the line models won't download (probably due to VPN or something IDK)
Abut I can manually clone repos
Ajust go in the model dir?
Four app auto crops
Fi improved qwen thing
Fnow it uses way lesser vram but you need to reinstall
Fhavent announced yet
Fwe use this qwen exactly let me show
FQwen/Qwen2.5-14B-Instruct
F
Fit is downloaded into hugging face cache folder
Fi will add it to the model downloader as option
Aty very much
Fsorry for delay
Fi am about to update
Fadding FPS increase
Z@Furkan Gözükara SECourses I downloaded the wan2.1 zip file, ran the massed compute, and then manually ran python Download_RIFE.py afterwards, and i'm still not seeing the updated fields in the gradio interface
Zi also followed the massed compute instructions and still not seeing the updated gradio fields
Fit was my mistake
Fwhen you reinstall
Fit will appear
Fi hadnt pushed update
Fso you need to reinstall or like update
Zi did reinstall, i deleted the previous wan2.1v11 folder and downloaded the v12, and reinstalled and still not there
ZZok it's working now
Fgreat
Fsorry for this issue i had forgatten to push
Mso is it normal for the wan2.1 i2v 480p model download to be 7 9GB files?
Cregaeding RIFE FPS increase , i have always error red color regarding sound at the end in CMD comsole. It is normall ?
AAfter much testing...
AI went back
Alol
AI just cannot make prodigy work as well for some reason
AIf I use ADAMW8Bit, ADAFactor, Lion
Aany of these I can get nice clean results
Aif I use Prodigy there's always some level of smeariness
Aand just very...weird non-compliance to prompts it's hard to explain. It's almost like selective overtraining - it passes my "blue hair" sampling test - but then it won't put the character into certain poses no matter what I try, for example
RThat's super interesting. Thank you for sharing your findings as well. I think I won't spend time experimenting with Prodigy for now. Did you notice any advantage in Adamw8bit or Lion over Adafactor?
RIf you get similar results and the training process requires less time, that's a win I would say.