You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a question: What is multi-allreduce? Why are there no performance tests for multi-allreduce in nccl-tests, and how can the performance of multi-allreduce be tested?
The text was updated successfully, but these errors were encountered:
I have a question: What is multi-allreduce? Why are there no performance tests for multi-allreduce in nccl-tests, and how can the performance of multi-allreduce be tested?
On two nodes, each with 8 GPUs, is multi-allreduce simply running multiple rank=16 allreduces simultaneously?
I have a question: What is multi-allreduce? Why are there no performance tests for multi-allreduce in nccl-tests, and how can the performance of multi-allreduce be tested?
The text was updated successfully, but these errors were encountered: