下载正常安装GraphVite不容易,如果你可以轻松安装,的确是运气很好。需要gcc版本,conda版本,机器环境刚好与作者相符合才行。
然后运行,基本的示例
graphvite baseline quick start
如果PyTorch版本太低,会报错
AttributeError: module 'torch' has no attribute ‘as_tensor'
如果PyTorch版本过高,会报错
The NVIDIA driver on your system is too old (found version 10000).
正常情况:
------------- link prediction --------------
effective edges: 6644 / 6650
effective filter edges: 327429 / 327429
remaining edges: 6644 / 6644
AUC: 0.902055
[time] GraphApplication.evaluate: 5.94148 s
macro-F1@20%: 0.248511
micro-F1@20%: 0.394835
[time] GraphApplication.evaluate: 26.2803 s
save embeddings and name mappings to `line_blogcatalog.pkl`
[time] GraphApplication.save: 0.0316839 s
可以见到以20%数据做预测,效果还是不错的,和论文基本一致。
数据是放在:xxx是你的user_name下面
/home/xxx/.graphvite/dataset/