pythonRuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing

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

• fork and spawn Are different ways to build subprocesses :

forkspawn Bifurcation lay eggs Inherit Not inherited Build a child process from scratch , The data of the parent process is copied to the space of the child process , Own your own Python Interpreter The variable has the same... As in the parent process idN Import modules at the beginning of each child process Subprocess get at name == main Variables defined in the block N

Solution

if __name__=='__main__':
torch.multiprocessing.set_start_method('spawn')

Reference and more

• The log says “RuntimeError: Cannot re-initialize CUDA in forked subprocess”
• https://britishgeologicalsurvey.github.io/science/python-forking-vs-spawn/
• https://www.pythonf.cn/read/65459
• https://stackoverflow.com/questions/64095876/multiprocessing-fork-vs-spawn