Python 为什么必须在 __name__ == '__main__' 子句中调用 multiprocessing.set_start_method？

Question

If you want to spawn a process, instead of fork a process, using multiprocess module, you can use the following command:

multiprocessing.set_start_method('spawn')

The documentation states:

To select a start method you use the set_start_method() in the if __name__ == '__main__' clause of the main module

However, I've found that I can call set_start_method('spawn') outside of the if __name__ == '__main__' clause, and it still spawns instead of forks processes on Unix.

In fact, as long as you invoke set_start_method('spawn') prior to the first invocation of multiprocessing.Process() , it will behave as expected. If you instead attempt to call set_start_method('spawn') after the first multiprocessing.Process() , it seems to be ignored and will fork instead of spawn.

Why does the documentation state that set_start_method should be invoked inside the if __name__ == '__main__' clause?

Answer 1

Sub-processes do not enter the if __name__ == '__main__' clause. That way we can guarantee set_start_method will be called only once.

It does not need to be specifically at the module level, but it can be in any function that is ran through that clause, and still guarantee only one invocation.