Skip to content

bound2020/EGES_multiGPU

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

参考了两位大佬的代码,简单实现了论文《Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba》中的EGES,并可以运行在单机多GPU上。

训练pipeline:

  1. 统计点击序列中的每个item pair的共现次数,保存为data/graph_node

  2. 生成每个item的side info,保存为data/side_info_feature

  3. 运行preprocess.py,根据图随机游走生成序列,保存为data/walk_seq

  4. 运行eges_multigpu.py,生成item向量

About

EGES多卡的简单实现

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages