From 2a2ad7b59abd09cb020f49f2774e382b307a29d1 Mon Sep 17 00:00:00 2001
From: frankwhzhang <frankwhzhang@126.com>
Date: Mon, 1 Jun 2020 13:53:32 +0800
Subject: [PATCH] mmoe readme

---
 models/multitask/mmoe/data/run.sh | 16 ++++++++++++++++
 models/multitask/mmoe/readme.md   | 13 +++++++++++++
 2 files changed, 29 insertions(+)
 create mode 100644 models/multitask/mmoe/data/run.sh

diff --git a/models/multitask/mmoe/data/run.sh b/models/multitask/mmoe/data/run.sh
new file mode 100644
index 00000000..b60d42b3
--- /dev/null
+++ b/models/multitask/mmoe/data/run.sh
@@ -0,0 +1,16 @@
+mkdir train_data
+mkdir test_data
+mkdir data
+train_path="data/census-income.data"
+test_path="data/census-income.test"
+train_data_path="train_data/"
+test_data_path="test_data/"
+pip install -r requirements.txt
+
+wget -P data/ https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census.tar.gz
+tar -zxvf data/census.tar.gz -C data/
+
+python data_preparation.py --train_path ${train_path} \
+                           --test_path ${test_path} \
+                           --train_data_path ${train_data_path}\
+                           --test_data_path ${test_data_path}
diff --git a/models/multitask/mmoe/readme.md b/models/multitask/mmoe/readme.md
index 4efa036a..694323db 100644
--- a/models/multitask/mmoe/readme.md
+++ b/models/multitask/mmoe/readme.md
@@ -14,6 +14,19 @@ python -m paddlerec.run -m paddlerec.models.multitask.mmoe
 
 根据原论文，我们在开源数据集Census-income Data上验证模型效果
 
+### 数据下载及预处理
+
+数据地址： [Census-income Data](https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census.tar.gz )
+
+数据解压后， 在data/run.sh脚本文件中添加文件的路径，并运行脚本。
+
+```shell
+cd data
+sh run.sh
+```
+
+脚本运行后，在config.yaml中修改数据路径dataset.data_path
+
 ### 参数
 
 config.yaml中的hyper_parameters部分，batch_size:32, epochs:400
-- 
GitLab