Datasets:

diffusers

/

dog-example

like 6

from huggingface_hub import snapshot_download

local_dir = "./dog"
snapshot_download(
"diffusers/dog-example",
local_dir=local_dir, repo_type="dataset",
ignore_patterns=".gitattributes",
)

export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export INSTANCE_DIR="dog"
export OUTPUT_DIR="path-to-save-model"

accelerate launch train_dreambooth.py
--pretrained_model_name_or_path=$MODEL_NAME
--instance_data_dir=$INSTANCE_DIR
--output_dir=$OUTPUT_DIR
--instance_prompt="a photo of sks dog"
--resolution=512
--train_batch_size=1
--gradient_accumulation_steps=1
--learning_rate=5e-6
--lr_scheduler="constant"
--lr_warmup_steps=0
--max_train_steps=400
--push_to_hub