I am trying to train a neural network under PyTorch. Everything works properly except when i specify in the dataloader num_workers=4 or even 3, it raises the error Error 11: Resource temporarily unavailable. I have a decent machine with 128 core CPU and a good gpu as well. I checked with htop to monitor the cpu usage and all the resources seem to be good. Checked with nvidia-smi for the gpu and also, nothing odd to report in terms of memory. Could you please help to troubleshoot this?
Do you see in the stacktrace which call fails and thus which resource might be unavailable?