UPDATE (11/11/2025) : for few days it was the wrong safetensors file for the qwen version, I've replaced it.

v1:
Trained both qwen & flux loras in a row on the same dataset.
The flux one is not that great, I may retrain it on a better dataset.