3 ms·
"Our biggest jobs run MPI, and all pods within the job are participating in a single MPI communicator." https://openai.com/research/scaling-kubernetes-to-7500-
by hackandthink 3y ago
"Our biggest jobs run MPI, and all pods within the job are participating in a single MPI communicator."
https://openai.com/research/scaling-kubernetes-to-7500-nodes https://openai.com/research/scaling-kubernetes-to-7500-nodes