Created by: wojtuss

A few days ago, when I was running a C-API inference application for MobileNet-SSD or any other model on CPU with a flag FLAGS_paddle_num_threads set to > 1, the inference run on multiple threads. With the current version of PaddlePaddle it does not work any more, i.e. the app runs only on a single thread. How should I set the number of cores to use in the C-API app now?