在使用graphsage的时候，有一个问题，如果use_id=True，结果就会变得很差，这个是什么原因造成的呢？

在使用graphsage模型的时候，把use_id改成True，不使用任何特征，发现结果基本上是随机的，并且save_embedding之后，发现embedding矩阵没有进行归一化

Start by reproducing the GraphSAGE configuration with use_id=True and no features, then inspect the embedding produced after save_embedding. Compare its behavior and normalization with the default feature-based configuration; done means the cause of the random results and the normalization behavior are identified and documented.

Written by the indexing model from the issue text.