SETTP: Style Extraction and Tunable Inference via Dual-level Transferable Prompt Learning | Read Paper on Bytez