OrionEdit-qwen

OrionEdit-qwen weights for cross-image editing, built on latest Qwen/Qwen-Image-Edit-2511.

This repo contains:

• pytorch_lora_weights.safetensors — LoRA on the diffusion transformer (rank 256)
• pytorch_processor_weights.safetensors — decoupling processors for attention blocks

Usage

Install the OrionEdit codebase, then load with:

from models.pipeline_orion_edit import OrionEditPipeline

pipeline = OrionEditPipeline.from_orion_pretrained(
    base_model="Qwen/Qwen-Image-Edit-2511",
    orion_repo="ZeyuJiang1/OrionEdit-qwen",
)