reddit 数据上，graphsage，scalable的模型速度问题

在wiki中看到，Reddit数据上，2层卷积，Graphsage的速度为0.12s/mini-batch, Scalable_gcn的速度为0.026s/mini-batch. 我自己用reddit数据在本地测试了一下，同样512大小的batch-size，graphsage的速度是0.37s/mini-batch， scalable_gcn的速度为0.34s/mini-batch. 差距比较大，想问一下在测速的时候还有什么其他设定吗？

The issue names no source file or test. Start by locating the Reddit benchmark and the GraphSAGE and Scalable_GCN timing setup, then compare the documented and local settings, including the 512 batch size. Done means identifying the setting differences or reproducing the reported timing gap.

Written by the indexing model from the issue text.