lingvo
lingvo copied to clipboard
low GPU utilization rate
In locally sync mode, the GPU utilization is unstable (sometime 100%, sometime 0%), and the average GPU utilization is about 60%. I think this question may caused by IO when data feeding. Any ways to solve it?
How the input generator work? Will it prefetching data for next batch when GPU is caculating current batch?
The input generator does not prefetch data so that might indeed be the cause.