fix error in dcn-mix (#326)

浅梦 · web-flow · commit 4762e851b160 · 2021-02-12T19:35:05.000+08:00
fix error in dcn-mix
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -36,6 +36,7 @@ jobs:
     - name: Install dependencies
       run: |
         pip3 install -q tensorflow==${{ matrix.tf-version }}
+        pip install -q numpy==1.19.5
         pip install -q requests
         pip install -e .
     - name: Test with pytest
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -7,14 +7,14 @@ If you
 - have spare time to learn and develop
 - familiar with git
 
-please send a brief introduction of your background and experience to wcshen1994@163.com, welcome to join us!
+please send a brief introduction of your background and experience to weichenswc@163.com, welcome to join us!
 
 # Creating a pull request
-1. **Become a collaborator**: Send an email with introduction and your github account name to wcshen1994@163.com and waiting for invitation to become a collaborator.
-2. **Fork&Dev**: Fork your own branch(`dev_yourname`) in `DeepCTR` from the `master` branch for development.If the `master` is updated during the development process, remember to merge and update to `dev_yourname` regularly.
-3. **Testing**: Test  logical correctness and effect when finishing the code development of the `dev_yourname` branch. 
-4. **Pre-release** : After testing contact wcshen1994@163.com for pre-release integration, usually your branch `dev_yourname` will be merged into `release` branch by squash merge. 
-5. **Release a new version**: After confirming that the change is no longer needed, `release` branch will be merged into `master` and a new python package will be released on pypi.
+1.  **Become a collaborator**: Send an email with introduction and your github account name to weichenswc@163.com and waiting for invitation to become a collaborator.
+2.  **Fork&Dev**: Fork your own branch(`dev_yourname`) in `DeepCTR` from the `master` branch for development.If the `master` is updated during the development process, remember to merge and update to `dev_yourname` regularly.
+3.  **Testing**: Test  logical correctness and effect when finishing the code development of the `dev_yourname` branch. 
+4.  **Pre-release** : After testing contact weichenswc@163.com for pre-release integration, usually your branch `dev_yourname` will be merged into `release` branch by squash merge. 
+5.  **Release a new version**: After confirming that the change is no longer needed, `release` branch will be merged into `master` and a new python package will be released on pypi.
 
 # Discussions
 
diff --git a/README.md b/README.md
@@ -13,7 +13,7 @@
 ![CI status](https://github.com/shenweichen/deepctr/workflows/CI/badge.svg)
 [![Coverage Status](https://coveralls.io/repos/github/shenweichen/DeepCTR/badge.svg?branch=master)](https://coveralls.io/github/shenweichen/DeepCTR?branch=master)
 [![Codacy Badge](https://api.codacy.com/project/badge/Grade/d4099734dc0e4bab91d332ead8c0bdd0)](https://www.codacy.com/app/wcshen1994/DeepCTR?utm_source=github.com&amp;utm_medium=referral&amp;utm_content=shenweichen/DeepCTR&amp;utm_campaign=Badge_Grade)
-[![Disscussion](https://img.shields.io/badge/chat-wechat-brightgreen?style=flat)](./README.md#disscussiongroup)
+[![Disscussion](https://img.shields.io/badge/chat-wechat-brightgreen?style=flat)](./README.md#DisscussionGroup)
 [![License](https://img.shields.io/github/license/shenweichen/deepctr.svg)](https://github.com/shenweichen/deepctr/blob/master/LICENSE)
 <!-- [![Gitter](https://badges.gitter.im/DeepCTR/community.svg)](https://gitter.im/DeepCTR/community?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge) -->
 
@@ -24,7 +24,7 @@ DeepCTR is a **Easy-to-use**,**Modular** and **Extendible** package of deep-lear
 - Provide  `tensorflow estimator` interface for **large scale data** and **distributed training**. [example](https://deepctr-doc.readthedocs.io/en/latest/Quick-Start.html#getting-started-4-steps-to-deepctr-estimator-with-tfrecord)
 - It is compatible with both `tf 1.x`  and `tf 2.x`.
 
-Some related project:
+Some related projects:
 - DeepMatch: https://github.com/shenweichen/DeepMatch
 - DeepCTR-Torch: https://github.com/shenweichen/DeepCTR-Torch
 
@@ -74,19 +74,11 @@ If you find this code useful in your research, please cite it using the followin
 }
 ```
 
-## DisscussionGroup 交流群
+## DisscussionGroup
 
 - [Discussions](https://github.com/shenweichen/DeepCTR/discussions)
 - 公众号：**浅梦的学习笔记**  
 - wechat ID: **deepctrbot**
 
   ![wechat](./docs/pics/code.png)
   
-##  Cooperative promotion 合作推广
-For more information about the recommendation system, such as **feature engineering, user profile, matching, ranking and multi-objective optimization, online learning and real-time computing, and more cutting-edge technologies and practical projects**:
-
-更多关于推荐系统的内容，如**特征工程，用户画像，召回，排序和多目标优化，在线学习与实时计算以及更多前沿技术和实战项目**等可参考：
-
-- [推荐系统实战](https://www.julyedu.com/course/getDetail/181?ccode=5ee751d37278c)
-- [互联网计算广告实战](https://www.julyedu.com/course/getDetail/158?ccode=5ee751d37278c)
-
diff --git a/deepctr/estimator/models/afm.py b/deepctr/estimator/models/afm.py
@@ -53,7 +53,7 @@ def _model_fn(features, labels, mode, config):
         linear_logits = get_linear_logit(features, linear_feature_columns, l2_reg_linear=l2_reg_linear)
 
         with variable_scope(DNN_SCOPE_NAME):
-            sparse_embedding_list, dense_value_list = input_from_feature_columns(features, dnn_feature_columns,
+            sparse_embedding_list, _ = input_from_feature_columns(features, dnn_feature_columns,
                                                                                  l2_reg_embedding=l2_reg_embedding)
             if use_attention:
 
diff --git a/deepctr/estimator/models/ccpm.py b/deepctr/estimator/models/ccpm.py
@@ -59,7 +59,7 @@ def _model_fn(features, labels, mode, config):
         linear_logits = get_linear_logit(features, linear_feature_columns, l2_reg_linear=l2_reg_linear)
 
         with variable_scope(DNN_SCOPE_NAME):
-            sparse_embedding_list, dense_value_list = input_from_feature_columns(features, dnn_feature_columns,
+            sparse_embedding_list, _ = input_from_feature_columns(features, dnn_feature_columns,
                                                                                  l2_reg_embedding=l2_reg_embedding)
             n = len(sparse_embedding_list)
             l = len(conv_filters)
diff --git a/deepctr/feature_column.py b/deepctr/feature_column.py
@@ -93,9 +93,9 @@ class DenseFeat(namedtuple('DenseFeat', ['name', 'dimension', 'dtype', 'transfor
         name: feature name,
         dimension: dimension of the feature, default = 1.
         dtype: dtype of the feature, default="float32".
-        transform_fn: If not None, a function that can be used to transfrom
+        transform_fn: If not `None` , a function that can be used to transform
         values of the feature.  the function takes the input Tensor as its
-        argument, and returns the output Tensor. 
+        argument, and returns the output Tensor.
         (e.g. lambda x: (x - 3.0) / 4.2).
     """
     __slots__ = ()
diff --git a/deepctr/layers/interaction.py b/deepctr/layers/interaction.py
@@ -399,14 +399,13 @@ def call(self, inputs, **kwargs):
             if self.parameterization == 'vector':
                 xl_w = tf.tensordot(x_l, self.kernels[i], axes=(1, 0))
                 dot_ = tf.matmul(x_0, xl_w)
-                x_l = dot_ + self.bias[i]
+                x_l = dot_ + self.bias[i] + x_l
             elif self.parameterization == 'matrix':
-                dot_ = tf.einsum('ij,bjk->bik', self.kernels[i], x_l)  # W * xi  (bs, dim, 1)
-                dot_ = dot_ + self.bias[i]  # W * xi + b
-                dot_ = x_0 * dot_  # x0 · (W * xi + b)  Hadamard-product
+                xl_w = tf.einsum('ij,bjk->bik', self.kernels[i], x_l)  # W * xi  (bs, dim, 1)
+                dot_ = xl_w + self.bias[i]  # W * xi + b
+                x_l = x_0 * dot_ + x_l  # x0 · (W * xi + b) +xl  Hadamard-product
             else:  # error
-                print("parameterization should be 'vector' or 'matrix'")
-            x_l = dot_ + x_l
+                raise ValueError("parameterization should be 'vector' or 'matrix'")
         x_l = tf.squeeze(x_l, axis=2)
         return x_l
 
diff --git a/deepctr/models/__init__.py b/deepctr/models/__init__.py
@@ -9,7 +9,6 @@
 from .fnn import FNN
 from .mlr import MLR
 from .onn import ONN
-from .onn import ONN as NFFM
 from .nfm import NFM
 from .pnn import PNN
 from .wdl import WDL
diff --git a/docs/source/Examples.md b/docs/source/Examples.md
@@ -367,7 +367,8 @@ if __name__ == "__main__":
                                          batch_size=2 ** 14, num_epochs=1, shuffle_factor=0)
 
     # 3.Define Model,train,predict and evaluate
-    model = DeepFMEstimator(linear_feature_columns, dnn_feature_columns, task='binary')
+    model = DeepFMEstimator(linear_feature_columns, dnn_feature_columns, task='binary', 
+                            config=tf.estimator.RunConfig(tf_random_seed=2021))
 
     model.train(train_model_input)
     eval_result = model.evaluate(test_model_input)
@@ -422,15 +423,16 @@ if __name__ == "__main__":
 
     # 3.generate input data for model
 
-    train, test = train_test_split(data, test_size=0.2, random_state=2020)
+    train, test = train_test_split(data, test_size=0.2, random_state=2021)
 
     # Not setting default value for continuous feature. filled with mean.
 
     train_model_input = input_fn_pandas(train, sparse_features + dense_features, 'label', shuffle=True)
     test_model_input = input_fn_pandas(test, sparse_features + dense_features, None, shuffle=False)
 
     # 4.Define Model,train,predict and evaluate
-    model = DeepFMEstimator(linear_feature_columns, dnn_feature_columns, task='binary')
+    model = DeepFMEstimator(linear_feature_columns, dnn_feature_columns, task='binary', 
+                            config=tf.estimator.RunConfig(tf_random_seed=2021))
 
     model.train(train_model_input)
     pred_ans_iter = model.predict(test_model_input)
diff --git a/docs/source/FAQ.md b/docs/source/FAQ.md
@@ -37,6 +37,18 @@ es = EarlyStopping(monitor='val_binary_crossentropy')
 history = model.fit(model_input, data[target].values,batch_size=256, epochs=10, verbose=2, validation_split=0.2,callbacks=[es] )
 ```
 
+If you are using Estimator models, you can set learning rate like:
+
+```python
+from deepctr.estimator import DeepFMEstimator
+import tensorflow as tf
+
+model = DeepFMEstimator(linear_feature_columns, dnn_feature_columns, task='binary', 
+                            linear_optimizer=tf.train.FtrlOptimizer(0.05), dnn_optimizer=tf.train.AdagradOptimizer(0.1)
+                            )
+
+```
+
 
 ## 3. Get the attentional weights of feature interactions in AFM
 --------------------------------------------------------------------------
diff --git a/docs/source/Features.md b/docs/source/Features.md
@@ -41,7 +41,7 @@ DNN based CTR prediction models usually have following 4 modules:
 - name : feature name
 - dimension : dimension of dense feature vector.
 - dtype : default `float32`.dtype of input tensor.
-- transform_fn : If not None, a function that can be used to transfrom
+- transform_fn : If not `None` , a function that can be used to transform
         values of the feature.  the function takes the input Tensor as its
         argument, and returns the output Tensor. 
         (e.g. `lambda x: (x - 3.0) / 4.2)`.
diff --git a/docs/source/History.md b/docs/source/History.md
@@ -1,4 +1,5 @@
 # History
+- 02/12/2021 : [v0.8.4](https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.4) released.Fix bug in DCN-Mix. 
 - 01/06/2021 : [v0.8.3](https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.3) released.Add [DCN-Mix](./Features.html#dcn-mix-improved-deep-cross-network-with-mix-of-experts-and-matrix-kernel) model.Support `transform_fn` in `DenseFeat`.
 - 10/11/2020 : [v0.8.2](https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.2) released.Refactor `DNN` Layer.
 - 09/12/2020 : [v0.8.1](https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.1) released.Improve the reproducibility & fix some bugs.
diff --git a/docs/source/index.rst b/docs/source/index.rst
@@ -42,31 +42,19 @@ You can read the latest code and related projects
 
 News
 -----
+02/12/2021 : Fix bug in DCN-Mix. `Changelog <https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.4>`_
+
 01/06/2021 : Add `DCN-Mix <./Features.html#dcn-mix-improved-deep-cross-network-with-mix-of-experts-and-matrix-kernel>`_ (`中文介绍 <https://mp.weixin.qq.com/s/0qidwbxyfTkODTw2DIiRWw>`_) and support ``transform_fn`` in ``DenseFeat``. `Changelog <https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.3>`_
 
 10/11/2020 : Refactor ``DNN`` Layer. `Changelog <https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.2>`_
 
-09/12/2020 : Improve the reproducibility & fix some bugs. `Changelog <https://github.com/shenweichen/DeepCTR/releases/tag/v0.8.1>`_
-
-DisscussionGroup 交流群
+DisscussionGroup
 -----------------------
 
 `Discussions <https://github.com/shenweichen/DeepCTR/discussions>`_  公众号：**浅梦的学习笔记**  wechat ID: **deepctrbot**
 
 .. image:: ../pics/code.png
 
-Cooperative promotion 合作推广
-------------------------------
-
-For more information about the recommendation system, such as **feature engineering, user profile, matching, ranking and multi-objective optimization, online learning and real-time computing, and more cutting-edge technologies and practical projects** :
-
-更多关于推荐系统的内容，如 **特征工程，用户画像，召回，排序和多目标优化，在线学习与实时计算以及更多前沿技术和实战项目** 等可参考：
-
-- `推荐系统实战 <https://www.julyedu.com/course/getDetail/181?ccode=5ee751d37278c>`_
-- `互联网计算广告实战 <https://www.julyedu.com/course/getDetail/158?ccode=5ee751d37278c>`_
-
-
-
 .. toctree::
    :maxdepth: 2
    :caption: Home:
diff --git a/examples/run_estimator_pandas_classification.py b/examples/run_estimator_pandas_classification.py
@@ -39,7 +39,7 @@
 
     # 3.generate input data for model
 
-    train, test = train_test_split(data, test_size=0.2, random_state=2020)
+    train, test = train_test_split(data, test_size=0.2, random_state=2021)
 
     # Not setting default value for continuous feature. filled with mean.
 
diff --git a/setup.py b/setup.py
@@ -11,7 +11,7 @@
     name="deepctr",
     version="0.8.3",
     author="Weichen Shen",
-    author_email="wcshen1994@163.com",
+    author_email="weichenswc@163.com",
     description="Easy-to-use,Modular and Extendible package of deep learning based CTR(Click Through Rate) prediction models with tensorflow 1.x and 2.x .",
     long_description=long_description,
     long_description_content_type="text/markdown",
diff --git a/tests/models/AutoInt_test.py b/tests/models/AutoInt_test.py
@@ -32,7 +32,6 @@ def test_AutoInt(att_layer_num, dnn_hidden_units, sparse_feature_num):
 def test_AutoIntEstimator(att_layer_num, dnn_hidden_units, sparse_feature_num):
     if not Estimator_TEST_TF1 and version.parse(tf.__version__) < version.parse('2.2.0'):
         return
-    model_name = "AutoInt"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/CCPM_test.py b/tests/models/CCPM_test.py
@@ -53,7 +53,6 @@ def test_CCPM_without_seq(sparse_feature_num, dense_feature_num):
 def test_CCPMEstimator_without_seq(sparse_feature_num, dense_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "CCPM"
 
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
diff --git a/tests/models/DCNMix_test.py b/tests/models/DCNMix_test.py
@@ -1,9 +1,7 @@
 import pytest
-import tensorflow as tf
 
 from deepctr.models import DCNMix
-from ..utils import check_model, get_test_data, SAMPLE_SIZE, get_test_data_estimator, check_estimator, \
-    Estimator_TEST_TF1
+from ..utils import check_model, get_test_data, SAMPLE_SIZE
 
 
 @pytest.mark.parametrize(
diff --git a/tests/models/DCN_test.py b/tests/models/DCN_test.py
@@ -33,7 +33,6 @@ def test_DCN(cross_num, hidden_size, sparse_feature_num, cross_parameterization)
 def test_DCNEstimator(cross_num, hidden_size, sparse_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "DCN"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/DeepFM_test.py b/tests/models/DeepFM_test.py
@@ -33,7 +33,6 @@ def test_DeepFM(hidden_size, sparse_feature_num):
 def test_DeepFMEstimator(hidden_size, sparse_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "DeepFM"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/FNN_test.py b/tests/models/FNN_test.py
@@ -47,7 +47,6 @@ def test_FNN(sparse_feature_num, dense_feature_num):
 def test_FNNEstimator(sparse_feature_num, dense_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "FNN"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/FiBiNET_test.py b/tests/models/FiBiNET_test.py
@@ -29,7 +29,6 @@ def test_FiBiNET(bilinear_type):
 def test_FiBiNETEstimator(bilinear_type):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "FiBiNET"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size, sparse_feature_num=2,
                                                                                     dense_feature_num=2)
diff --git a/tests/models/FwFM_test.py b/tests/models/FwFM_test.py
@@ -31,7 +31,6 @@ def test_FwFM(hidden_size, sparse_feature_num):
 def test_FwFMEstimator(hidden_size, sparse_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "FwFM"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/NFM_test.py b/tests/models/NFM_test.py
@@ -29,7 +29,6 @@ def test_NFM(hidden_size, sparse_feature_num):
 def test_FNNEstimator(hidden_size, sparse_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "NFM"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/PNN_test.py b/tests/models/PNN_test.py
@@ -29,7 +29,6 @@ def test_PNN(use_inner, use_outter, sparse_feature_num):
 def test_PNNEstimator(use_inner, use_outter, sparse_feature_num):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "PNN"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
diff --git a/tests/models/WDL_test.py b/tests/models/WDL_test.py
@@ -19,7 +19,7 @@ def test_WDL(sparse_feature_num, dense_feature_num):
     model_name = "WDL"
     sample_size = SAMPLE_SIZE
     x, y, feature_columns = get_test_data(sample_size, sparse_feature_num=sparse_feature_num,
-                                          dense_feature_num=dense_feature_num)
+                                          dense_feature_num=dense_feature_num, hash_flag=True)
 
     model = WDL(feature_columns, feature_columns,
                 dnn_hidden_units=[4, 4], dnn_dropout=0.5)
@@ -34,7 +34,6 @@ def test_WDL(sparse_feature_num, dense_feature_num):
 def test_WDLEstimator(sparse_feature_num, dense_feature_num):
     if not Estimator_TEST_TF1 and version.parse(tf.__version__) < version.parse('2.2.0'):
         return
-    model_name = "WDL"
     sample_size = SAMPLE_SIZE
 
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size, sparse_feature_num,
diff --git a/tests/models/xDeepFM_test.py b/tests/models/xDeepFM_test.py
@@ -48,7 +48,6 @@ def test_xDeepFMEstimator(dnn_hidden_units, cin_layer_size, cin_split_half, cin_
                           dense_feature_dim):
     if not Estimator_TEST_TF1 and tf.__version__ < "2.2.0":
         return
-    model_name = "xDeepFM"
     sample_size = SAMPLE_SIZE
     linear_feature_columns, dnn_feature_columns, input_fn = get_test_data_estimator(sample_size,
                                                                                     sparse_feature_num=sparse_feature_num,
