optimum-intel-ldm-super-resolution-4x

This repo provides a simple ldm super resolution example of how to use Optimum-intel to optimize and accelerate inference of Hugging Face Model Ldm-super-resolution with OpenVINO on Intel CPU.

Installation

• optimum-intel==1.5.2 (include openvino)
  • openvino
  • openvino-dev
• diffusers
• pytorch >= 1.9.1
• onnx >= 1.13.0

Setup Environment

# Install optimum-intel 
conda create -n optimum-intel python=3.8
conda activate optimum-intel
python -m pip install -r requirements.txt

# Install diffusers 
# by pip
pip install --upgrade diffusers[torch]
# by conda 
conda install -c conda-forge diffusers

HuggingFace Ldm super resolution pipeline

python HF-ldm4x-pipeline.py

OpenVINO Ldm model convert to IR

python ov-ldm4x-model-convert.py

OpenVINO Ldm super resolution pipeline

python ov-ldm4x-pipeline.py

ToDo

• ldm super resolution model int8 quantization by NNCF