qwen-edit-turbo v7: bake Qwen4Play_v2 lora (off by default, lora_qwen4play toggles), drop Amateur-Photo lora; start.sh passes no VRAM flag (full-coverage loras OOM under --highvram; perf unchanged). Image: comfyui-serverless:qwen-edit-turbo-v7

serverless: RunPod worker (handler, build/deploy scripts, Dockerfile, qwen-edit-turbo endpoint)