kalle07/qwen2.1-image-turbo-lora-set

This is a LoRa extraction from Qwen-Image-2.1 and Qwen-Image-2.1-Turbo model.
using KJ nodes from comfyui

I have created a series of LoRa use cases for the base model with 6–10 steps (CGF 1).

3 Ranks 32/64/128 all bf16

32 rank is smalles and reveals minor differences

64 and 128 very similar (and extreme close to original 178 rank from comfyui)

source of other qwen turbo loras (for me better in img2img)

https://huggingface.co/congruency/Qwen-Image-2.1-turbo-mfd