Hello Qwen team,

I’m building a commercial SaaS product that involves AI-based image generation and editing. We’re currently testing Qwen2-VL and Qwen-Image-Edit as potential base models because of their strong image-text alignment and style consistency.

We saw that the Qwen architecture supports LoRA fine-tuning, but official training scripts for LoRA or adapter integration aren’t yet released.

Could you please confirm:

1. Whether there’s an internal or upcoming LoRA training script release planned;
2. If not, whether you can share the recommended way to inject LoRA adapters into Qwen2-VL or Qwen-Image-Edit manually (e.g., supported target_modules or layer names);
3. If there’s a contact or Slack/Discord channel where we can follow LoRA-related updates for Qwen models.

Thank you for your excellent open models — we’d love to build on Qwen for creative AI applications.