for 30 images how many images in my classification dataset and how much "class images per instance
for 30 images how many images in my classification dataset and how much "class images per instance images" should i set it to?
M1/1 - for one classification image one instance image. No more need. Use Furkan's reg images.
MAfter the recent Prodigy testing, I have now styled with AdamW. Using the same image set. So far the Prodigy has won.
MHere is my config for style. I used 10 images with 10 repeats, captions and 10 epoch
Fwith dreambooth extension
Fif class images are good quality
Fi suggest 100 per instance
Fso you would need 3000
Li will use your class images
Fok make it 100
Lok so i set it to 100 and use 3000 images
thank you
thank you
L@Dr. Furkan Gözükara my dataset is 512px x 768px and I have trained max resolution 768px. Any particupar reason why the generated images have more ressemblance with my dataset when generate at 768px x 768px compared to those generated 512px x 768px?
GHi all, I'm a brand new Patreon supporter. I'm following the tutorial to make a LoRA training of myself in Kaggle. But I get the following error:
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 14.76 GiB total capacity; 13.17 GiB already allocated; 7.75 MiB free; 13.46 GiB reserved in total by PyTorch Here is my training command: accelerate launch --num_cpu_threads_per_process=2 "./sdxl_train_network.py" --pretrained_model_name_or_path="stabilityai/stable-diffusion-xl-base-1.0" --train_data_dir="/kaggle/working/results/img" --reg_data_dir="/kaggle/working/results/reg" --resolution="1024,1024" --output_dir="/kaggle/working/results/model" --logging_dir="/kaggle/working/results/log" --network_alpha="1" --save_model_as=safetensors --network_module=networks.lora --text_encoder_lr=0.0004 --unet_lr=0.0004 --network_dim=32 --output_name="kaggle_glpr123" --lr_scheduler_num_cycles="8" --no_half_vae --learning_rate="0.0004" --lr_scheduler="constant" --train_batch_size="1" --max_train_steps="6400" --save_every_n_epochs="1" --mixed_precision="fp16" --save_precision="fp16" --cache_latents --optimizer_type="Adafactor" --optimizer_args scale_parameter=False relative_step=False warmup_init=False --max_data_loader_n_workers="0" --bucket_reso_steps=64 --full_fp16 --xformers --bucket_no_upscale --noise_offset=0.0 --lowram --max_grad_norm=0.0MIf I'm tagging with WD14 and it's a drawing style, but I don't want it to be an asian/manga type model, which one should I train on?
MMeinaMix maybe?
Fwhich model? also did you enable bucketing?
Fhello welcome
Fgradient checkpointing missing.
Fi think you need to test on models
AHi, it is possible to use own lora in rerender?
Aand is it possible to make the first keyframe separately? (made in the automation interface, for example)
Fi think yes
Fi am not sure but probably possible
Fit is open source
DReddit
Explore this post and more from the StableDiffusion community
Ftesting righ tnow
Fso many packages to install lol
Flol still installing
Fit failed haha
Dohhh
Fthe interface is also terrible
Fsd 1.5 worked
Falmost same speed
LI am using Photon_v1 as based model, not sure what is bucketing
Deverytime i want to do a 32:9 aspect ratio wallpaper , it struggles and spits a runtime error. anyone had any luck ? I run dual monitor and thought 2240x630 would work..
Dthen i can upscale , i dont think it works that way
Deven the width is limited to 2048
EWhat kind of class classification can i use for training my child (3 years old) - like man\woman(child?)
Dhttps://github.com/cheald/sd-webui-loractl
This extension sounds that we can balance loras depending on how powerful they are. Very interesting. (thanks to reddit)
This extension sounds that we can balance loras depending on how powerful they are. Very interesting. (thanks to reddit)
GitHub
An Automatic1111 extension for dynamically controlling the weights of LoRAs during image generation - GitHub - cheald/sd-webui-loractl: An Automatic1111 extension for dynamically controlling the we...
L@Dr. Furkan Gözükara do you think thwt your auto cropping script can help to solve it?
D@Dr. Furkan Gözükara , is your tutorial on controlnet same for SDXL too ? Only difference being the SDXL models right ?
Mwho can tell me to reinstall Windows, I run vlad sdnext, and it doesn't start, what to do
Ahow can i use controlnet in sdxl in runpod? It only installs outdated models
AI tried to update canny but it does not initialize controlonet
Dhey , relly i cant find on youtube , a updated workflow . haw to make my own videous into animation with stable diffusion , @Dr. Furkan Gözükara you made a video like 6 mounths ago , could you please point me in a updated direction , i dont need perfection i just want to create my ai video creatios , my daughters into someghing
EI wonder if we could use this amazing sdxl checkpoint to train our characters dreambooth - https://learn.thinkdiffusion.com/thinkdiffusionxl-model-draft/
ADeleted the download folder in /working, deleted cn-models in the root folder, still copied and pasted all over again, no way I can not update
MI've been looking at Lora's profiles for days, trying to find the best settings for poses and styles. I also trained two Lora's today, and in one of them, epoch 9 and epoch 10 produce completely different output with the same seed, so there are some interesting switches to hash out.
I'm curious, for either DB or Lora, what do you use to set the number of iterations of the folders? Because I see illogical iteration numbers like 12 or 16, while the number of images is 104 and similar non-round numbers. So just by taste, maybe there is a rule?
I'm curious, for either DB or Lora, what do you use to set the number of iterations of the folders? Because I see illogical iteration numbers like 12 or 16, while the number of images is 104 and similar non-round numbers. So just by taste, maybe there is a rule?