About cpu when load image and num_workers

cc5b4cb40e77805f1eec · January 13, 2019, 2:45am

I have the following two questions

I cannot set the num_workers in Dataloader to any number grater then 0. it will have an error like . what’s the problem?

RuntimeError: DataLoader worker (pid 28830) is killed by signal: Bus error. Details are lost due to multiprocessing. Rerunning with num_workers=0 may give better error trace

when num_workers = 0, the cpu(6 cores, 12 threads) seems to be running with multiple threads… why?

ranklord · January 13, 2019, 7:02am

Reasons could be running this in a docker container, in which case there is not enough shared memory. Or simply leave this parameter away Check this thread, for instance: Give a better error when we run out of shared memory, instead of "RuntimeError: DataLoader worker (pid 13) is killed by signal: Bus error." · Issue #5040 · pytorch/pytorch · GitHub

cc5b4cb40e77805f1eec · January 14, 2019, 10:35am

Thanks for replying!
Actually not in a docker container…in a ubuntu server.I will try to expand the shared memory .