От пяти сломанных версий трансформеров до 4,45 секунд на плате: внедрение vision tower Qwen2-VL-2B на RK3588

Оригинал

From five broken transformers versions to 4.45 seconds on the board: getting Qwen2-VL-2B's vision tower onto an RK3588

Two days ago this model wouldn't even export. Today its vision tower runs on a real RK3588 board: a 392×392 image encoded in 4,445.68 ms (median of three independent processes), model load about 7.3 seconds, zero RKNN errors at runtime. This is a field note from the conversion chain — what broke, what the fix actually was, and what the number does and doesn't mean. Where it started: the official demo doesn't run on the transformers it targets Rockchip's official RKLLM multimodal demo