# AutoInt

**Repository Path**: magicat128/auto-int

## Basic Information

- **Project Name**: AutoInt
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-05-13
- **Last Updated**: 2021-05-13

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# AutoInt

This is a TenforFlow implementation of AutoInt on Ascend 910 for CTR prediction task, as described in paper:

Weiping Song, Chence Shi, Zhiping Xiao, Zhijian Duan, Yewen Xu, Ming Zhang and Jian Tang. [AutoInt: Automatic Feature Interaction Learning via Self-Attentive Neural Networks](https://arxiv.org/pdf/1810.11921.pdf). arXiv preprint arXiv:1810.11921, 2018.

This implementation achieves around 0.8400 AUC and 0.3828 Logloss on the MovieLens-1M dataset. The original paper reports 0.8456 AUC and 0.3797 Logloss.

Dataset available at: https://pan.baidu.com/s/1bq3ZaK0uVXFLe5Gt-z2mzA 

Password: gi1k 

## Requirements

- Tensorflow 1.15
- Python 3.7
- Ascend 910 (Image path: swr.cn-north-4.myhuaweicloud.com/ascend-share/3.3.0.alpha002_tensorflow-ascend910-cp37-euleros2.8-aarch64-training:1.15.0-2.0.12_0412)

## Usage

```
python3 train.py \
	--data_url PATH_TO_DATA \
	--train_url PATH_TO_OUTPUT \
	--epoch 30 \
	--batch_size 1024 \
	--learning_rate 0.001 \
	--field_size 7
```