Shakker-Labs/AWPortrait-QW
Adapter vs Qwen/Qwen-Image. Shakker-Labs/AWPortrait-QW is a LoRA adapter, not full weights: rank 32, on 0, 1, 2, add_k_proj, add_q_proj, add_v_proj, to_add_out, to_k… (720 adapted matrices). It modifies Qwen/Qwen-Image at load time.
1 other repos hold byte-for-byte identical weight files.