Repositios Khoya interesante
Kohya SS Gui Repo : https://github.com/bmaltais/kohya_ss
Aqui estan los comandos para instalar Khoya en Runpod y los comandos para descargar los modelos desde civitai
Realistic Vision V2 Full model : https://huggingface.co/SG161222/Realistic_Vision_V2.0/resolve/main/Realistic_Vision_V2.0.safetensors
wget https://huggingface.co/SG161222/Realistic_Vision_V2.0/resolve/main/Realistic_Vision_V2.0.safetensors
RPG V4 Full model: https://huggingface.co/danbrown/RPG-v4/tree/main
wget https://huggingface.co/danbrown/RPG-v4/resolve/main/rpg-v4.safetensors
wget https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.ckpt
git clone https://github.com/bmaltais/kohya_ss.git
cd kohya_ss
python3 -m venv venv
source venv/bin/activate
apt update
apt-get install python3.10-tk
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
./setup.sh -n
unset LD_LIBRARY_PATH
bash gui.sh --share
cd /workspace && git clone https://github.com/bmaltais/kohya_ss.git && cd kohya_ss && python3 -m venv venv && source venv/bin/activate && apt update -y && apt-get install python3.10-tk -y && pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 && unset LD_LIBRARY_PATH && ./setup.sh -n
root@2d6e404b7b8d:/workspace/stable-diffusion-webui/models# rsync -rtlv ../../training/model/ ./Lora/
Parametro | Valor |
---|---|
Train batch size | 2 |
Epoch | 1 |
Save every epoch | 1 |
Caption extension | .txt |
Mixed precision | bf16 |
Save precision | bf16 |
number of CPU threats per core | 2 |
seed | 1234 |
learning rate | 0.001 |
LR scheduler | constant |
LR warmup (% of steps) | 10 |
Cache laten | marcado |
text encoder learning rate | 0,00005 |
unet learning rate | 0.0001 |
network rank | 128 |
max resolution | 512,512 |
stop text encoder training | 0 |
enable buckets | marcado |
max resolution | 512,512 |
max resolution | 512,512 |
Advance configuration | |
Use xformers | desmarcado |
Skip clip | 2 |