CAUTION, this project currently is experimental!
Currently you have to prepare pretrained models and datasets manually.
Download Huggingface models via git lfs, eg.
git clone https://huggingface.co/runwayml/stable-diffusion-v1-5# (optional) prepare your python environment,
# recommanded conda with miniforge
pip install -U -r requirements_dev.txt
# (Optional) if bitsandbytes caused error,
# you should add this env
export LD_LIBRARY_PATH=/home/xxx/miniforge3/envs/xxx/lib/python3.10/site-packages/nvidia/cuda_runtime/lib:$LD_LIBRARY_LATH
# step 1. prepare your dataset
# eg.
git clone https://yourdataset /path/to/dataset
# step 2. prepare your config file: train.yaml
cp train.yaml your_conf.yaml
vim your_conf.yaml
# step 3. start your training
./mobfusion -f your_conf.yaml- Add
dadaptationsupport - Add Lora/Lycoris support