How to properly use

[Fighterrr007]

Dear author, can you update the readme with the code on how to use it specifically? For example, I want to ablate the concept of violence, what should I do?

[nupurkmr9]

Hi, you can find the directions to use our code using diffusers here.
The specific commands to use for ablating the concept of violence is also given there and copied here for reference:

accelerate config
export MODEL_NAME="CompVis/stable-diffusion-v1-4"
export OUTPUT_DIR="logs_ablation/violence"

## launch training script (2 GPUs recommended, if 1 GPU increase --max_train_steps to 800 or increase --train_batch_size=8)

accelerate launch train.py \
          --pretrained_model_name_or_path=$MODEL_NAME  \
          --output_dir=$OUTPUT_DIR \
          --class_data_dir=./data/samples_violence/ \
          --class_prompt="people"  \
          --caption_target "violent, horrifying" \
          --concept_type violence \
          --resolution=512  \
          --train_batch_size=4  \
          --learning_rate=2e-7  \
          --max_train_steps=400 \
          --scale_lr --hflip \
          --parameter_group full-weight \
          --enable_xformers_memory_efficient_attention 

Hope this helps. Thanks!