[Bug/Discussion] Unable to load VAE from qwen_image_edit_2511_fp8_e4m3fn_scaled_lightning.safetensors (VAE is invalid: None)

[YellowDragonLive]

Description (描述)
I am trying to use the new Qwen-Image-Edit-2511 FP8 Lightning model (all-in-one safetensors, approx. 20GB). While the model and CLIP seem to load, the VAE connection returns None, and I get the error: VAE is invalid: None.

I am using an RTX 3060 Ti (8GB VRAM) with 32GB System RAM.

Steps to Reproduce (重现步骤)
Download qwen_image_edit_2511_fp8_e4m3fn_scaled_lightning.safetensors.

Use Load Checkpoint or Qwen2VLModelLoader node to load the file.

Connect the VAE output to a VAE Decode node.

Prompt the model and queue the prompt.

Error occurs: VAE is invalid: None.

Expected Behavior (预期行为)
The single .safetensors file should contain the VAE weights and allow the VAE Decode node to function without needing an external VAE file.

Actual Behavior (实际行为)
The VAE component is not recognized within the single file, or fails to initialize on 8GB VRAM hardware, resulting in a NoneType error.

Environment (环境信息)
OS: Windows 10/11

GPU: NVIDIA RTX 3060 Ti (8GB)

RAM: 32GB (Pagefile set to 60GB)

ComfyUI Version: [填写你的版本，如 v0.3.50]

Relevant Custom Nodes: ComfyUI-Kijai / ComfyUI-GGUF

[comfyui-wiki]

It's not a docs-related report, and you are not using the correct model, not using the correct model loader

And it's a diffusion model, not VAE. Please check this page for Qwen-Image-Edit 2511 support: https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit-2511

The lightv2x team just created a new model that can work in ComfyUI. Check it here: https://huggingface.co/lightx2v/Qwen-Image-Edit-2511-Lightning/blob/main/qwen_image_edit_2511_fp8_e4m3fn_scaled_lightning_comfyui.safetensors