runpod recently having huge issues
runpod recently having huge issues
Ftesting installers now
Nfor real
F
Fdmn this speed
F@RMK ask that question to @NicB
R@NicB Could you guys create a way to access JupyterLab inside MasseCompute from an external machine?
For example, JupyterLab is currently on a localhost URL.
Can you make a public URL so that I can access it from an outside network computer like an outside Windows or Mac?
For example, JupyterLab is currently on a localhost URL.
Can you make a public URL so that I can access it from an outside network computer like an outside Windows or Mac?
Fi am fixing the error
Nyou are the best Dr for real.
Fthanks
Fupdated and fixed
FFi tested adetailer and reactor
F
F@Dr. Furkan Gözükara which face + clothes method is best now? ip adapter 2?
F(no training)
V@Dr. Furkan Gözükara yes the same question, which face + clothes extension is the best, ip adapter 2?
Finstant id best for face
Fclothing idm vton
Fface instant id clothing idm vton
Fwow training of kohya way slower
Fthan before
Fon runpod
Vanything better than idm vton so far?
Ffor clothing transfer best
Fstarted 9 kohya training
F
Fnot clothes change. i mean having a face (just a face) and a dress, and compose image
Fin idm vton you need a full body picture right?
Fwtf
Fmulti gpu training?
Fhow much time it takes to train on 9x?
For 9 different trainings
Fdoes idm vton work with just a face?
Fwithout body*
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dno
Dwell it swaps
D
Dbut the results are like
Flol
F9 different
Fi do train batch size 1 to find best
R@Dr. Furkan Gözükara what is the current best dreambooth training for kohya? You've done a lot of tests. Is it the 48gb_textencoder.json in your https://www.patreon.com/posts/very-best-for-of-89213064?
Fnot updated yet
Fcurrently working to update
Rcool. You're updating the TE training version of db adafactor ?
Ryou have it at TE lr = 1/3 Unet lr
RYoure testing for betterTE training?
RFi am testing both
FTE and Unet right now