DeepMatch-Torch

DeepMatch-Torch is a PyTorch version of DeepMatch.

DeepMatch-Torch is a PyTorch Version deep matching model library for recommendations & advertising. It's easy to train models and to export representation vectors for user and item which can be used for ANN search. You can use any complex model with model.fit() and model.predict(). And you can keep same habit of using DeepMatch.

Let's Get Started! or Run examples !

1. Install

1. Install deepmatch-torch by pip

pip install deepmatch-torch

2. Install deepmatch-torch through source code

git clone https://github.com/bbruceyuan/DeepMatch-Torch.git
cd DeepMatch-Torch
python setup.py install

2. Tutorial

You can run example by following steps.

cd examples
python run_youtubednn.py or run_fm_dssm.py

3. Models List

Model	Paper
FM	[ICDM 2010]Factorization Machines
DSSM	[CIKM 2013]Deep Structured Semantic Models for Web Search using Clickthrough Data
YoutubeDNN	[RecSys 2016]Deep Neural Networks for YouTube Recommendations
NCF	[WWW 2017]Neural Collaborative Filtering
MIND	[CIKM 2019]Multi-interest network with dynamic routing for recommendation at Tmall

TODO

• simplify model config. now only support kwargs, but config is a elegant choice.
• fix MIND only support CPU train bug.

Acknowledgments

Especially thanks to DeepMatch. This project relies highly on DeepMatch. Additionally, I used the PLBaseModel design from torchTS.

Thanks to this awesome projects.

中文 README

DeepMatch-Torch 是一个 PyTorch 版本的 DeepMatch。

DeepMatch-Torch 是一个用于广告推荐的召回模型库，可以非常简单地训练模型和导致 user 和 item 的 vertor 表示，你可以用这个 user/item 的表示进行 ANN 近似检索。在 DeepMatch-Torch 中，你可以保持和 DeepMatch 一样的习惯，通过 model.fit() 和 model.predict() 进行模型的训练和预测。

更多的中文文档细节可以参见：中文文档

1. 安装

• 步骤一: 安装 PyTorch, 按照官网指引
• 步骤二：
  • 通过 pip 安装：pip install deepmatch-torch
  • 通过源码安装

git clone https://github.com/bbruceyuan/DeepMatch-Torch.git
cd DeepMatch-Torch
python setup.py install

2. 快速上手

进入 example 目录查看相关代码，使用方式和 DeepMatch 几乎保持一致。

cd examples
python run_youtubednn.py or run_fm_dssm.py