Oppo-A5S-Firmware

<div align="center"> <h1>Qwen-Image-Edit-Pruning</h1> <a href='https://arxiv.org/abs/2511.16156'><img src='https://img.shields.io/badge/arXiv-2508.07607-b31b1b.svg'></a> &nbsp; <a href='https://github.com/OPPO-Mente-Lab/Qwen-Image-Pruning'><img src="https://img.shields.io/badge/GitHub-OPPOer-blue.svg?logo=github" alt="GitHub"></a> </div>

Update

• 2025/10/09: We release Qwen-Image-Edit-2509-Pruning-13B-4steps
• 2025/09/29: We release Qwen-Image-Edit-2509-Pruning-14B
• 2025/09/28: We release Qwen-Image-Edit-Pruning-13B-4steps

Introduction

This open-source project is based on Qwen-Image-Edit and has attempted model pruning, removing 20 layers while retaining the weights of 40 layers, resulting in a model size of 13.6B parameters. The pruned version will continue to be iterated upon. Please stay tuned.

<div align="center"> <img src="bench-2509.png"> </div>

Quick Start

Install the latest version of diffusers and pytorch

pip install torch
pip install git+https://github.com/huggingface/diffusers

Qwen-Image-Edit-2509-14B Inference

import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
model_name = f"OPPOer/Qwen-Image-Edit-2509-Pruning"
pipeline = QwenImageEditPlusPipeline.from_pretrained(model_name, torch_dtype=torch.bfloat16)
print("pipeline loaded")
pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)
image1 = Image.open("input1.jpg")
image2 = Image.open("input2.jpg")
prompt = "Let the ancient costume beauty in the second picture sit on the sofa in the first picture"
inputs = {
    "image": [image1, image2],
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}
with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_plus.png")
    print("image saved at", os.path.abspath("output_image_edit_plus.png"))

Qwen-Image-Edit-2509-13B Inference

import os
import torch
from PIL import Image
from diffusers import QwenImageEditPlusPipeline
model_name = f"OPPOer/Qwen-Image-Edit-2509-Pruning/Qwen-Image-Edit-2509-13B-4steps"
pipeline = QwenImageEditPlusPipeline.from_pretrained(model_name, torch_dtype=torch.bfloat16)
print("pipeline loaded")
pipeline.to('cuda')
pipeline.set_progress_bar_config(disable=None)
image1 = Image.open("input1.jpg")
image2 = Image.open("input2.jpg")
prompt = "Let the ancient costume beauty in the second picture sit on the sofa in the first picture"
inputs = {
    "image": [image1, image2],
    "prompt": prompt,
    "generator": torch.manual_seed(0),
    "true_cfg_scale": 1.0,
    "negative_prompt": " ",
    "num_inference_steps": 4,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}
with torch.inference_mode():
    output = pipeline(**inputs)
    output_image = output.images[0]
    output_image.save("output_image_edit_plus.png")
    print("image saved at", os.path.abspath("output_image_edit_plus.png"))

Citation

🌟 If you find our work helpful, please consider citing our paper and leaving valuable stars

@misc{ma2025pluggablepruningcontiguouslayer,
      title={Pluggable Pruning with Contiguous Layer Distillation for Diffusion Transformers}, 
      author={Jian Ma and Qirong Peng and Xujie Zhu and Peixing Xie and Chen Chen and Haonan Lu},
      year={2025},
      eprint={2511.16156},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2511.16156}, 
}