3 ms·Scaling TensorFlow inference to unlimited items per request with bounded latency1 points by RealJon 8y ago