recommend-tf-estimator

Models List

Model	Paper
DeepFM	[arXiv 2017] DeepFM: A Factorization-Machine based Neural Network for CTR Prediction
ESMM	[SIGIR 2018] Entire Space Multi-Task Model: An Effective Approach for Estimating Post-Click Conversion Rate
MMOE	[KDD 2018] Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts
FiBiNET	[RecSys 2019] FiBiNET: Combining Feature Importance and Bilinear feature Interaction for Click-Through Rate Prediction
TwoTower	[arXiv 2020] Embedding-based Retrieval in Facebook Search

code sturcture

--config            训练配置，可根据业务新增 
--data              数据样本 
--src 
    --input_fn      输入相关函数 
    --models_ompl   模型实现
    --common_utils  通用工具函数，包含特殊层，特殊loss的实现
--examples          运行样例
--online_deploy     部署脚本
--test              测试脚本

--common_utils  
layers                  特殊层实现(SENet, 双线性交叉层, attention层等)
loss_fn                 损失函数
wpai_model_auto_update  更新在线预测的模型

--layers
dice
prelu
build_deep_layers
build_Bilinear_Interaction_layers
build_SENET_layers
attention_layer
batch_norm_layer

quick start

注意：数据和特征需要自己定义和添加到代码中

cd examples
python train_esmm.py

* mmoe 实现了base和wide+esmm版本
使用方法：
在初始化estimaor时，指定model_fn即可