hello.I have found throughput metric in nodes/s in some papers.What does that mean?Does this metric mean dividing the input nodes of each epoch by the total epoch time? If so, does it have the same meaning as epoch time? I use simple neighbor sampling and run it on a distributed platform.
throughput may indicates different metric which could be data copy, sampling in different papers. whether this metric mean dividing input nodes of each epoch depends on the paper context. Here’s the paper introducing DistDGL(throughput is mentioned) and you could refer to it.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.