JoyFusionAI/Qwen-Image-2.1-Uncensored-LoRA

Qwen-Image-2.1 Uncensored LoRA (mirror)

Mirror of qwen-image-2.1-uncensored-lora.safetensors from abenzerps/Qwen-Image-2.1-Uncensored-GGUF, hosted for JoyFusion clients. Not modified.

• Base model: Qwen/Qwen-Image-2.1 (7B single-stream DiT). Does not work with the older 20B Qwen-Image models.
• Format: PEFT (transformer_blocks.N.attn.{to_q,to_k,to_v,to_out.0}.lora_{A,B}.default.weight), rank 16, attention projections only, 256 tensors, bf16.
• sha256: 00ee2cb7f00a37b097640c1880ed0d17e368e30599f34bb348fee0e8c2cc4ca1
• Loads in mflux ≥ 0db68699 (mflux-generate-qwen-2.1 --lora-paths ... --lora-scales 1.0) and in Draw Things ≥ v26.0924.0 via LoRA import.
• License follows the base model: Qwen Research License (non-commercial).