awesome-semantic-segmentation-pytorch

awesome-semantic-segmentation-pytorch copied to clipboard

If I interrupt multi-GPU training, sometimes there will be several zombie processes. How can I avoid this situation?

Which model and backbone are you using for multi-gpu training?

Which model and backbone are you using for multi-gpu training?

Any one

For the time being, I dont have a work around. However, most often the first process is the main process. If you kill that one, other zombie process seems to die.