Thank you for your suggestion. I'm re - modifying the main model for multi - GPU retraining.
Thank you for your suggestion. I'm re - modifying the main model for multi - GPU retraining.
Ffor multi gpu loras works great with 48 gb gpus
Fdreambooth need 80 gb gpus
筋4X 4090 48G
Fawesome
Fi wish they were sold worldwide
Fdefinitely better than 32 gb 5090
TTried training character checkpoint with 50 image set 7000 steps from flux krea and result is much worse than flux dev counterpart so far. Much less likeness, seems more weird results
Fi feel krea need more epochs
KI found using a character Lora that I trained with flux on flux krea is giving me the textures of krea with the character. Vs the finetune I was getting fuzzy images
Tis a 80gb gpu needed for flux krea Dreambooth/Finetune?
As far as I remember for flux dev DB/Finetune a 48gb gpu was enough
As far as I remember for flux dev DB/Finetune a 48gb gpu was enough
F6 gb sufficient
Fif you want multiple gpu yes
ZHi,
I was trying to generate using my trained dreambooth model on Runpod. It shown this error in SwarmUI. When I run it in massedcompute, it's working fine. Do you what is the cause of this error?
Thank you
I was trying to generate using my trained dreambooth model on Runpod. It shown this error in SwarmUI. When I run it in massedcompute, it's working fine. Do you what is the cause of this error?
Thank you
Fmodel download failed
Frm -r modelpath
Fredownload
S@Dr. Furkan Gözükara Hey Dr, is there a theoratical limit with checkpoint training regarding dataset ? I have a good 350 images, is that too much or I would just need to train it longer ?
Uuisato_
Instagram
⟆ - tʀɪße_∞ : [2]
regⱭlia∴CΥΛN∴bøne⟲loop∴ritual
ƒ͓*c͓κ͓ †͓h͓ɘ͓ ͓α͓l͓ĝ͓ø͓ // attn=∅ ✱
#animation #cinema #design #tribal
regⱭlia∴CΥΛN∴bøne⟲loop∴ritual
ƒ͓*c͓κ͓ †͓h͓ɘ͓ ͓α͓l͓ĝ͓ø͓ // attn=∅ ✱
#animation #cinema #design #tribal
Likes
371
Swan2.2?
EHi Furkan, I want to confirm something about your captioning theory.
For a small and very consistent dataset (around 50 solid product photos on white background), is it still best practice to only use the token + class as caption (like CCCSNOOO bag), and let the images carry all visual attributes?
Or in such small homogeneous datasets, could adding more detailed captions (e.g. color, material, shape) actually help stabilize training with T5?
For a small and very consistent dataset (around 50 solid product photos on white background), is it still best practice to only use the token + class as caption (like CCCSNOOO bag), and let the images carry all visual attributes?
Or in such small homogeneous datasets, could adding more detailed captions (e.g. color, material, shape) actually help stabilize training with T5?
JGood question! AFAIk depends on what you want. iI you dont caption a certain thing, it will always generate it in that thing, unless you really force it in a prompt.
If you caption something (like color, material) you can more easily change the color , material later.
I think Furkan will confirm this
If you caption something (like color, material) you can more easily change the color , material later.
I think Furkan will confirm this
EThanks for the clarification! If T5 is active with such a small and homogeneous dataset, could relying only on token + class captions cause it to fall back to it's pretraining bias (e.g. anime/fantasy)? In that case, would the safer option be to disable T5?
Also, around what epoch should the results typically start to become visible? I’ve tested the token + class approach, but even after 190 epochs I still mostly get strange anime-like outputs. Does that indicate that the LoRA simply needs more training, or that the captioning strategy itself is unsuitable for this setup?
Also, around what epoch should the results typically start to become visible? I’ve tested the token + class approach, but even after 190 epochs I still mostly get strange anime-like outputs. Does that indicate that the LoRA simply needs more training, or that the captioning strategy itself is unsuitable for this setup?
JI have no idea about T5, is that Clip?
What model are you training on? SDXL, Pony , Flux, Qwen, Wan 2.1, Wan 2.2?
What model are you training on? SDXL, Pony , Flux, Qwen, Wan 2.1, Wan 2.2?
JWhat's your learning rate
EFlux. It's this config setup.
JAre you able to generate images while training?
ENo, but I have three computers.
JI think your learning rate is extremely low
Jnow I havent trained Flux, but 5e-05 is pretty low
Jmeaning 0.00005
Jtry learning rate 2e-4
EYes, 5e-5 looks low compared to typical SDXL LoRAs. But for Flux1 it’s the standard recommendation (e.g. Furkan’s config). It trains slower, but keeps things stable since Flux is more sensitive than SDXL/1.5. But maybe I'm wrong.
JYeah I am coming mostly from SD 1.5 , SDXL and Wan
Ji still would try going for 2e-4 with flux and see if you get decent results around 1500-3000 steps
JWorth a shot if you are not happy with your results
Jtry to find out if you can generate images while it's training (that means , the training algorithm stops for a moment to generate an image based on a prompt set)
Jso every 100 steps generate two images, one with prompt identical to a training image, and another one with something you may want to prompt later
Jthe later is more important
JYes , samples 
EThank you very much!
PGuys, can you help? I don't understand why the image is not properly generated... What am i missing? I don't understand how wan works...
Aplay with the return_noise and add_noise
Atheres something wrong there
PThanks for the advice - i used the same setting from base workflow text to video - but the result didn't change much
Atry 4+4 steps
Jtry this workflow
JJonXL Multi Wan 2.1 + 2.2 Advanced V5 AIO Workflow - T2I T2V I2V FLF2V (Trim, Extend, Upscale, Interpolate) * V5 released! Now supports Wan 2.1 wit...
Jits better if you use the 4 steps Lora
JDownload links are in the workflows' notes
PThanks, i definitely will! Wan is a new territory for me 