3 ms·Inference is latency sensitive so the front end is still relevant.by mirker 4y agoInference is latency sensitive so the front end is still relevant.