I want to use multiple nodes in Pytorch. How do I enter the ip addresses of different devices? I do not see any instructions on the tutorials page: http://pytorch.org/tutorials/intermediate/dist_tuto.html, which handles the case of a single machine only.
if you submit you script on different machines, and specify the master IP address, port and world size, this will work
Hi, thanks for quick reply…while launching scipts from different nodes, do I need to mention different ranks, say from node 1 I mention rank = 1, and from node I mention rank = 2?
yes, you need to specify the rank for each scripts
thanks for reply. Is there a way where I can run the script on a single machine, and specify the ip address of all the machines in that script? That would be very convenient when using many machines.
to the best of my knowledge, you must specify the script with the same master IP address, and submit you script on different machines. you can ssh to the target IP machines and submit you script.