/stable_diffusion_quantizer.pytorch

Custom pipelines for stable diffusion model

Primary LanguagePythonApache License 2.0Apache-2.0

Stable Diffusion Quantizer

!!!Work In Progress!!!

PyTorch implementation of Stable Diffusion quantization pipeline

Requirements

  • Python 3.8+
  • 1–8 high-end NVIDIA GPUs with at least 24 GB of memory.

Training

pip install -r requirements.txt
python quantize.py --models-dir <path_to_diffusers_sd_checkpoint> --cfg <path_to_quantizer_cfg> --gpus <n_gpus>