用deepwalk训练，loss和mrr值一直不怎么变

请问：
单机版，用deepwalk训练，loss和mrr值一直不怎么变是怎么回事啊？
一样的代码，我用python3写了下，效果好很多。
mrr这个评估值，是有效的吗？

The report does not name any files, tests, or entry points. Start by locating the single-machine DeepWalk training and MRR evaluation paths, then reproduce the unchanged loss and MRR and compare them with the reported Python 3 implementation. Done means identifying whether the behavior is expected or correcting it with evidence that the metrics respond appropriately.

Written by the indexing model from the issue text.