update document and add demo yaml

Author: tangzhiyi11

models/rank/naml/config_kunlun.yaml

@@ -0,0 +1,48 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+runner:
+  train_data_dir: "data/sample_data/train"
+  train_reader_path: "NAMLDataReader" # importlib format
+  use_gpu: False
+  use_xpu: True
+  train_batch_size: 10
+  epochs: 20
+  print_interval: 2
+  #model_init_path: "output_model/0" # init model
+  model_save_path: "output_model_all"
+  infer_batch_size: 2
+  infer_reader_path: "NAMLDataReader" # importlib format
+  test_data_dir: "data/sample_data/train"
+  infer_load_path: "output_model_all"
+  infer_start_epoch: 0
+  infer_end_epoch: 3
+
+# hyper parameters of user-defined network
+hyper_parameters:
+  # optimizer config
+  optimizer:
+    class: Adam
+    learning_rate: 0.001
+    strategy: async
+  # user-defined <key, value> pairs
+  article_content_size: 30
+  article_title_size: 10
+  browse_size: 10
+  neg_condidate_sample_size: 4
+  word_dimension: 30
+  category_size: 4
+  sub_category_size: 10
+  category_dimension: 32
+  word_dict_size: 101

models/rank/naml/train_on_kunlun.md

@@ -1,33 +1,56 @@
-# How to train naml on kunlun
+# 使用昆仑XPU芯片加速NAML模型训练
 
-## Prepare kunlun environment
-[Paddle installation for machines with Kunlun XPU card](https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/2.0-rc1/install/install_Kunlun_zh.html)
+## 准备Paddle昆仑XPU版训练环境
+[昆仑XPU芯片运行飞桨](https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/xpu_docs/index_cn.html)
 
-## Prepare data
+## 数据准备
+
+### 示例数据
+参考 [数据准备](README##数据准备)
+
+
+### 全量数据
 ```shell
 cd PaddleRec/datasets/MIND/data
 bash run.sh
 ```
 
-## Train
+## 训练
 ```shell
-# set kunlun card id
+# 设置训练使用的昆仑XPU芯片卡号
 export FLAGS_selected_xpus=0
-# enable convolution autotune
+# 开启昆仑XPU芯片卷积计算加速(可不设置)
 export XPU_CONV_AUTOTUNE=2
 
 cd PaddleRec/models/rank/naml 
-python3.7 -u ../../../tools/trainer.py -m config_bigdata_kunlun.yaml
+# 全量数据动态图训练
+python3.7 -u ../../../tools/trainer.py -m config_bigdata_kunlun.yaml # 使用示例数据，请指定config_kunlun.yaml
+# 全量数据静态图训练
+python3.7 -u ../../../tools/static_trainer.py -m config_bigdata_kunlun.yaml # 使用示例数据，请指定config_kunlun.yaml
 ```
 
-
-## Eval
+## 评估
 ```shell
-# set kunlun card id
+# 设置训练使用的昆仑XPU芯片卡号
 export FLAGS_selected_xpus=0
-# enable convolution autotune
+# 开启昆仑XPU芯片卷积计算加速(可不设置)
 export XPU_CONV_AUTOTUNE=2
 
 cd PaddleRec/models/rank/naml 
-python3.7 -u ../../../tools/infer.py -m config_bigdata_kunlun.yaml
+# 全量数据动态图预测
+python3.7 -u ../../../tools/infer.py -m config_bigdata_kunlun.yaml # 使用示例数据，请指定config_kunlun.yaml
+# 全量数据静态图预测
+python3.7 -u ../../../tools/static_infer.py -m config_bigdata_kunlun.yaml # 使用示例数据，请指定config_kunlun.yaml
 ```
+
+## 模型效果
+以下为全量数据训练2个epoch的结果:
+
+| 模型 | 训练auc |batch_size | epoch_num| Time of each epoch| 
+| :------| :------ | :------ | :------| :------ | 
+| naml | 0.71 | 50 | 2 | 约7小时 | 
+
+
+| 模型 | 预测auc |batch_size | Time of each epoch| 
+| :------| :------ | :------ | :------ | 
+| naml | 0.67 | 10 | 约2小时 |