Fi plan to test it actually
Fsince multi gpu for fine tune fails :/
Feven on 48 gb
Fcurrently i locked windows to 2.5 RC
Fi think it works fine
Mim not sure if its the new cudnns or the torch versions, but i noticed massive quality and learning drop off
Fcan totally happen
Mi am going to test as well rn 
M@Dr. Furkan Gözükara can having different cuda toolkit affect results?
Fi doubt that
Fi seen only different xformers in past
Fbroken training
Fso spda can do same
Myou're right i just tested and sdpa gives worse results
F48 gb multi gpu still fails :/
FFLUX: Gradient checkpointing enabled. CPU offload: True
number of trainable parameters: 11901408320
number of trainable parameters: 11901408320
Ftried latest accelerator and torch
Dwhy do you offload if you have 48 gb?
MPlease let me know if this isn't the right place for questions... I followed the tutorial for Swarm UI and to get Flux into Swarm UI. During the installation process I did have an error which resolved when I turned off my Nord VPN connection.
I am able to generate images just fine using Stable Diffusion 3, but when I try either flux model, I receive the following error:
Self-Start ComfyUI-0 on port 7821 failed. Restarting per configuration AutoRestart=true...
[BackendHandler] backend #0 failed to load model with error: System.Net.WebSockets.WebSocketException (0x80004005): The remote party closed the WebSocket connection without completing the close handshake.
Anyone know how to resolve? I've made sure both the Nord VPN app is disconnected and the Nord browser extension as well.
I am able to generate images just fine using Stable Diffusion 3, but when I try either flux model, I receive the following error:
Self-Start ComfyUI-0 on port 7821 failed. Restarting per configuration AutoRestart=true...
[BackendHandler] backend #0 failed to load model with error: System.Net.WebSockets.WebSocketException (0x80004005): The remote party closed the WebSocket connection without completing the close handshake.
Anyone know how to resolve? I've made sure both the Nord VPN app is disconnected and the Nord browser extension as well.
LHello, can you tell me how to reduce the scales of a trained lora flux? At my exit, lora is weighing 1gb
Z@Dr. Furkan Gözükara can models do like different font styles? or maybe do like font wildcards for making alogo?
Fmulti gpu doesnt work even with offload
Fmulti gpu traning overheard
Fhello welcome
Fswarmui developer wants you to ask in their discord
FFplease ask there
Fother than that it will reduce quality
Flora network rank 64 32 16
Fi think flux can do yes
D@Dr. Furkan Gözükara koha training is running much faster with the new version
Dthere is some news about this?
LOne user wrote to me that my workout has problems due to the weight of 1gb. For some reason, his models are 18 mb each. I have seen such models with a similar weight, but they did not bring me high-quality results and differences
Fsingle layer
Freduces quality
Fthey dont test anything properly to compare
Fwhich version? torch 2.5?
Fyou can use lora resize of kohya and see quality difference
Lthanks
Z@Dr. Furkan Gözükara on supir app, if guidance scale needs to be 2, where do i set that? Text guidance scale?
DI'm not sure, I updated to train individual layers but with not a good success, now I training all layers and is super fast
D
Dresolution? bs? card?
D
D1024x1024
D
D
D1024x1024 rtx 3090
D
