right now i'm using clip-l vit-l-14-text and t5xxl_enconly.safetensors and it's messing up text
right now i'm using clip-l vit-l-14-text and t5xxl_enconly.safetensors and it's messing up text
Fi use fp16 t5 and default clip l for flux
F
DI found it learnt a mfking mosquito in one photo lol
MHello. Which setting sets the clip l learning rate in the config? Text encoder learning rate?
Fyes text encoder learning rate
F
N@Dr. Furkan Gözükara DR is there a problem with supir?
N
Nno sorry my bad.. I tried it again on a different instance for the 2nd time and it worked
Nsorry for the pin @
TTraceback (most recent call last):
File "E:\OneTrainer\OneTrainer\modules\ui\TrainUI.py", line 552, in __training_thread_function
trainer.start()
File "E:\OneTrainer\OneTrainer\modules\trainer\GenericTrainer.py", line 122, in start
self.model = self.model_loader.load(
File "E:\OneTrainer\OneTrainer\modules\modelLoader\FluxLoRAModelLoader.py", line 46, in load
base_model_loader.load(model, model_type, model_names, weight_dtypes)
File "E:\OneTrainer\OneTrainer\modules\modelLoader\flux\FluxModelLoader.py", line 261, in load
raise Exception("could not load model: " + model_names.base_model)
Exception: could not load model: E:/stable-diffusion-webui-forge/webui_forge_cu121_torch231/webui/models/Stable-diffusion/flux1-dev.safetensors
File "E:\OneTrainer\OneTrainer\modules\ui\TrainUI.py", line 552, in __training_thread_function
trainer.start()
File "E:\OneTrainer\OneTrainer\modules\trainer\GenericTrainer.py", line 122, in start
self.model = self.model_loader.load(
File "E:\OneTrainer\OneTrainer\modules\modelLoader\FluxLoRAModelLoader.py", line 46, in load
base_model_loader.load(model, model_type, model_names, weight_dtypes)
File "E:\OneTrainer\OneTrainer\modules\modelLoader\flux\FluxModelLoader.py", line 261, in load
raise Exception("could not load model: " + model_names.base_model)
Exception: could not load model: E:/stable-diffusion-webui-forge/webui_forge_cu121_torch231/webui/models/Stable-diffusion/flux1-dev.safetensors
TI have this model, but when I start training it says there is no model @Dr. Furkan Gözükara
PIs there any nice Lora for fixing hands and faces in Flux you guys can recommend? Especially when the character is further away form the cam i get issues .
P
Xthis looks amazing
Xyes I've noticed also smiling a lot in my generations. Also face tends to be in the middle of frame usually
�I would like to know about this also! One solution that I've found works well and is helpful especially if you know how to draw is to generate a Canny preview and then download and edit the preview in Photoshop. So pretty much erase the lines of the hand and redraw the basic details and then reupload it into Controlnet and that will guide Flux to fill in the rest.
PThats a nice trick.. I need to do 60+ images for a client.. that might not be in the budget hahha but good trick 
. I try atm to use Boreal Lora, and <segment:face,0.7> . But with later i am not sure if it works just like that in SD Forge or i need to add specific add on 
?
. I try atm to use Boreal Lora, and <segment:face,0.7> . But with later i am not sure if it works just like that in SD Forge or i need to add specific add on ?
Adr 256 images?!
Aall the same outfit? or various
Fawesome
Fonetrainer auto downloads
Fyou have to use hugging face repo
Fi did another training time to test :d
Fvarious but not all
Fdifferent times as well
Ni thought someone shared some flux style presets but I can find it - ring a bell for anyone?
SDr. Furkan I hope everything is going well, I wanted to ask you if I wanted to train the Arcane style from the Netflix series with Flux, how many images would be necessary? And in the process of doing so, the only thing that would change would be the class, right? I would change to style.
Fmore better
Fflux so powerful i think just style may work too
Flike ohwx style
For arcane style
Fimproving the app
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
DMust have been fighting Mr Freeze
AI made a lora of my wife and shes wearing the same clothes, is it better for her to wear different outfits? it seems like her clothes are baked into all the prompts
Fdifferent is better
Fif you dont want same clothing
Ahave you had luck having anyone hold a flash light
Aor literally anything like a shovel or broom or anything
Fnever tried
FI decided to be lazy and skip making all images square for a Flux dataset and turned on bucketing instead. The results are still really good. This was a dataset with a character and a Wiimote
FI tried the same with SDXL and it was a complete disaster. SDXL could not learn what a Wiimote is or how to hold one
Xdo loras get better the more they are used?
VHello
VI'm pretty much new to this self-hosted AI generation, I've been working for the last week with Fooocus (using SDXL) and I manage to build a create some consistent images of my AI Influencer.
Fflux lora is amazing
VThen I came across Flux and I kinda fell for it, but I've been having a hardtime doing the swapfaces feature like I used to do on Fooocus(that automatically generates based on the prompt and a set of faces you provide as reference).
Fcompared to sdxl
VNow I`m using SwarmUI, but I have no clue on hoe to achieve the faces I have already with future generations. Does anyone have a suggestion of how can I go about this?