uWSGI +Flask+torch Cannot re-initialize CUDA in forked subprocess

用uWSGI +Flask多进程多线程部署你的深度学习应用时，访问会报这样的错

"Cannot re-initialize CUDA in forked subprocess. To use CUDA with "
RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method

查了半天，终于查到怎么使用spawn方法而不是fork

在uWSGI的配置文件中，添加 lazy=true 即可