c++ - 如何阻止 TensorFlow 多线程

Question

我正在为 NIST FRVT 编写代码。NIST 希望程序最多运行 2 个线程(只有 CPU，没有 GPU)。我在我的代码中使用了 TensorFlow，但它总是产生 2 个以上的线程。 I tried this solution. It decreased the number of threads, but not up to 2

我收到这个警告

[WARNING] We've detected that your software may be threading or using other multiprocessing techniques during template creation. The number of threads detected was 9 and it should be 2. Per the API document, implementations must run single-threaded. In the test environment, there is no advantage to threading, because NIST will distribute workload across multiple blades and multiple processes. We highly recommend that you fix this issue prior to submission.

NIST 通过 top -H -b -n1 | 计算线程grep 验证 11 | wc -l

有没有办法强制 TensorFlow 使用最多 2 个线程？

是否有可以在 2 个线程上运行的 TensorFlow 版本？

(因为是TensorFlow，我通过去掉TensorFlow的部分代码来检查)

tensorflow 版本 1.8.0

opencv 版本 3.4.1

g++ 版本 4.8.5

g++ -std=c++11

Answer 1

引用以下资料，似乎不可能在 1 或 2 个线程上运行 TensorFlow。

• https://github.com/tensorflow/tensorflow/issues/33627
• https://github.com/tensorflow/tensorflow/issues/42510
• https://github.com/theislab/batchglm/issues/27
• Using the script variable OMP_NUM_THREADS in the program source files
• Change number of threads for Tensorflow inference with C API

我提交了具有 6 个以上线程的 NIST-FRVT 实现，NIST 接受了。 6 个线程中只有 1 个在运行，其余线程处于 sleep 模式。