I found that when batchsize is large, set shuffle=False will speed a lot indgl.dataloading.NodeDataLoader.
Is there any method to speed nodedataloader except to set shuffle=False ?
I benchmarked with OGB products on an AWS g3.16x instance and found that the speed difference is not too big.
Could you share your benchmarking code and hardware settings?
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.