update EDCN &fix typo

shenweichen · Nov 9, 2022 · 91dc7c6 · 91dc7c6
1 parent c13aba6
commit 91dc7c6
Show file tree

Hide file tree

Showing 23 changed files with 309 additions and 218 deletions.
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -1,6 +1,6 @@
-name: CI
+name: CI_TF2
 
-on: 
+on:
   push:
     path:
       - 'deepctr/*'
@@ -9,17 +9,17 @@ on:
     path:
       - 'deepctr/*'
       - 'tests/*'
-      
+
 jobs:
   build:
 
     runs-on: ubuntu-latest
     timeout-minutes: 180
     strategy:
       matrix:
-        python-version: [3.6,3.7,3.8,3.9,3.10.7]
-        tf-version: [1.4.0,1.15.0,2.6.0,2.7.0,2.8.0,2.9.0,2.10.0]
-        
+        python-version: [ 3.6,3.7,3.8, 3.9,3.10.7 ]
+        tf-version: [ 2.6.0,2.7.0,2.8.0,2.9.0,2.10.0 ]
+
         exclude:
           - python-version: 3.7
             tf-version: 1.4.0
@@ -64,31 +64,31 @@ jobs:
           - python-version: 3.10.7
             tf-version: 2.7.0
     steps:
-
-    - uses: actions/checkout@v3
-
-    - name: Setup python environment
-      uses: actions/setup-python@v4
-      with:
-        python-version: ${{ matrix.python-version }}
 
-    - name: Install dependencies
-      run: |
-        pip3 install -q tensorflow==${{ matrix.tf-version }}
-        pip install -q protobuf==3.19.0
-        pip install -q requests
-        pip install -e .
-    - name: Test with pytest
-      timeout-minutes: 180
-      run: |
-        pip install -q pytest
-        pip install -q pytest-cov
-        pip install -q python-coveralls
-        pytest --cov=deepctr --cov-report=xml
-    - name: Upload coverage to Codecov  
-      uses: codecov/[email protected]
-      with:
-        token: ${{secrets.CODECOV_TOKEN}}
-        file: ./coverage.xml
-        flags: pytest
-        name: py${{ matrix.python-version }}-tf${{ matrix.tf-version }}
+      - uses: actions/checkout@v3
+
+      - name: Setup python environment
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        run: |
+          pip3 install -q tensorflow==${{ matrix.tf-version }}
+          pip install -q protobuf==3.19.0
+          pip install -q requests
+          pip install -e .
+      - name: Test with pytest
+        timeout-minutes: 180
+        run: |
+          pip install -q pytest
+          pip install -q pytest-cov
+          pip install -q python-coveralls
+          pytest --cov=deepctr --cov-report=xml
+      - name: Upload coverage to Codecov
+        uses: codecov/[email protected]
+        with:
+          token: ${{secrets.CODECOV_TOKEN}}
+          file: ./coverage.xml
+          flags: pytest
+          name: py${{ matrix.python-version }}-tf${{ matrix.tf-version }}
diff --git a/.github/workflows/ci2.yml b/.github/workflows/ci2.yml
@@ -0,0 +1,96 @@
+name: CI_TF1
+
+on:
+  push:
+    path:
+      - 'deepctr/*'
+      - 'tests/*'
+  pull_request:
+    path:
+      - 'deepctr/*'
+      - 'tests/*'
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+    timeout-minutes: 180
+    strategy:
+      matrix:
+        python-version: [ 3.6,3.7 ]
+        tf-version: [ 1.15.0 ]
+
+        exclude:
+          - python-version: 3.7
+            tf-version: 1.4.0
+          - python-version: 3.7
+            tf-version: 1.12.0
+          - python-version: 3.7
+            tf-version: 1.15.0
+          - python-version: 3.8
+            tf-version: 1.4.0
+          - python-version: 3.8
+            tf-version: 1.14.0
+          - python-version: 3.8
+            tf-version: 1.15.0
+          - python-version: 3.6
+            tf-version: 2.7.0
+          - python-version: 3.6
+            tf-version: 2.8.0
+          - python-version: 3.6
+            tf-version: 2.9.0
+          - python-version: 3.6
+            tf-version: 2.10.0
+          - python-version: 3.9
+            tf-version: 1.4.0
+          - python-version: 3.9
+            tf-version: 1.15.0
+          - python-version: 3.9
+            tf-version: 2.2.0
+          - python-version: 3.9
+            tf-version: 2.5.0
+          - python-version: 3.9
+            tf-version: 2.6.0
+          - python-version: 3.9
+            tf-version: 2.7.0
+          - python-version: 3.10.7
+            tf-version: 1.4.0
+          - python-version: 3.10.7
+            tf-version: 1.15.0
+          - python-version: 3.10.7
+            tf-version: 2.2.0
+          - python-version: 3.10.7
+            tf-version: 2.5.0
+          - python-version: 3.10.7
+            tf-version: 2.6.0
+          - python-version: 3.10.7
+            tf-version: 2.7.0
+    steps:
+
+      - uses: actions/checkout@v3
+
+      - name: Setup python environment
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install dependencies
+        run: |
+          pip3 install -q tensorflow==${{ matrix.tf-version }}
+          pip install -q protobuf==3.19.0
+          pip install -q requests
+          pip install -e .
+      - name: Test with pytest
+        timeout-minutes: 180
+        run: |
+          pip install -q pytest
+          pip install -q pytest-cov
+          pip install -q python-coveralls
+          pytest --cov=deepctr --cov-report=xml
+      - name: Upload coverage to Codecov
+        uses: codecov/[email protected]
+        with:
+          token: ${{secrets.CODECOV_TOKEN}}
+          file: ./coverage.xml
+          flags: pytest
+          name: py${{ matrix.python-version }}-tf${{ matrix.tf-version }}
diff --git a/deepctr/__init__.py b/deepctr/__init__.py
@@ -1,4 +1,4 @@
 from .utils import check_version
 
-__version__ = '0.9.2'
+__version__ = '0.9.3'
 check_version(__version__)
diff --git a/deepctr/layers/__init__.py b/deepctr/layers/__init__.py
@@ -1,17 +1,16 @@
 import tensorflow as tf
 
 from .activation import Dice
-from .core import DNN, LocalActivationUnit, PredictionLayer, RegulationLayer
+from .core import DNN, LocalActivationUnit, PredictionLayer, RegulationModule
 from .interaction import (CIN, FM, AFMLayer, BiInteractionPooling, CrossNet, CrossNetMix,
                           InnerProductLayer, InteractingLayer,
                           OutterProductLayer, FGCNNLayer, SENETLayer, BilinearInteraction,
-                          FieldWiseBiInteraction, FwFMLayer, FEFMLayer, BridgeLayer)
+                          FieldWiseBiInteraction, FwFMLayer, FEFMLayer, BridgeModule)
 from .normalization import LayerNormalization
 from .sequence import (AttentionSequencePoolingLayer, BiasEncoding, BiLSTM,
                        KMaxPooling, SequencePoolingLayer, WeightedSequenceLayer,
-                       Transformer, DynamicGRU,PositionEncoding)
-
-from .utils import NoMask, Hash, Linear, _Add, combined_dnn_input, softmax, reduce_sum
+                       Transformer, DynamicGRU, PositionEncoding)
+from .utils import NoMask, Hash, Linear, _Add, combined_dnn_input, softmax, reduce_sum, Concat
 
 custom_objects = {'tf': tf,
                   'InnerProductLayer': InnerProductLayer,
@@ -28,7 +27,6 @@
                   'SequencePoolingLayer': SequencePoolingLayer,
                   'AttentionSequencePoolingLayer': AttentionSequencePoolingLayer,
                   'CIN': CIN,
-                  'RegulationLayer': RegulationLayer,
                   'InteractingLayer': InteractingLayer,
                   'LayerNormalization': LayerNormalization,
                   'BiLSTM': BiLSTM,
@@ -39,6 +37,7 @@
                   'FGCNNLayer': FGCNNLayer,
                   'Hash': Hash,
                   'Linear': Linear,
+                  'Concat': Concat,
                   'DynamicGRU': DynamicGRU,
                   'SENETLayer': SENETLayer,
                   'BilinearInteraction': BilinearInteraction,
@@ -50,5 +49,6 @@
                   'FEFMLayer': FEFMLayer,
                   'reduce_sum': reduce_sum,
                   'PositionEncoding': PositionEncoding,
-                  'BridgeLayer': BridgeLayer
+                  'RegulationModule': RegulationModule,
+                  'BridgeModule': BridgeModule
                   }
diff --git a/deepctr/layers/core.py b/deepctr/layers/core.py
@@ -267,41 +267,39 @@ def get_config(self, ):
         return dict(list(base_config.items()) + list(config.items()))
 
 
-class RegulationLayer(Layer):
+class RegulationModule(Layer):
     """Regulation module used in EDCN.
 
       Input shape
-        - A list of 3D tensor with shape: ``(batch_size,1,embedding_size)``.
+        - 3D tensor with shape: ``(batch_size,field_size,embedding_size)``.
 
       Output shape
-        - 2D tensor with shape: ``(batch_size, embedding_size * field_num)``.
+        - 2D tensor with shape: ``(batch_size,field_size * embedding_size)``.
 
       Arguments
         - **tau** : Positive float, the temperature coefficient to control
         distribution of field-wise gating unit.
 
-        - **seed** : A Python integer to use as random seed.
-
       References
         - [Enhancing Explicit and Implicit Feature Interactions via Information Sharing for Parallel Deep CTR Models.](https://dlp-kdd.github.io/assets/pdf/DLP-KDD_2021_paper_12.pdf)
     """
 
-    def __init__(self, tau=0.1, **kwargs):
+    def __init__(self, tau=1.0, **kwargs):
         if tau == 0:
-            raise ValueError("RegulationLayer tau can not be zero.")
+            raise ValueError("RegulationModule tau can not be zero.")
         self.tau = 1.0 / tau
-        super(RegulationLayer, self).__init__(**kwargs)
+        super(RegulationModule, self).__init__(**kwargs)
 
     def build(self, input_shape):
-        self.field_num = int(input_shape[1])
+        self.field_size = int(input_shape[1])
         self.embedding_size = int(input_shape[2])
         self.g = self.add_weight(
-            shape=(1, self.field_num, 1),
+            shape=(1, self.field_size, 1),
             initializer=Ones(),
             name=self.name + '_field_weight')
 
         # Be sure to call this somewhere!
-        super(RegulationLayer, self).build(input_shape)
+        super(RegulationModule, self).build(input_shape)
 
     def call(self, inputs, **kwargs):
 
@@ -311,13 +309,13 @@ def call(self, inputs, **kwargs):
 
         feild_gating_score = tf.nn.softmax(self.g * self.tau, 1)
         E = inputs * feild_gating_score
-        return tf.reshape(E, [-1, self.field_num * self.embedding_size])
+        return tf.reshape(E, [-1, self.field_size * self.embedding_size])
 
     def compute_output_shape(self, input_shape):
-        return (None, self.field_num * self.embedding_size)
+        return (None, self.field_size * self.embedding_size)
 
     def get_config(self):
         config = {'tau': self.tau}
-        base_config = super(RegulationLayer, self).get_config()
+        base_config = super(RegulationModule, self).get_config()
         base_config.update(config)
         return base_config
diff --git a/deepctr/layers/interaction.py b/deepctr/layers/interaction.py
@@ -1493,48 +1493,45 @@ def get_config(self):
         return config
 
 
-class BridgeLayer(Layer):  # ridge
-    """AttentionPoolingLayer layer used in EDCN
+class BridgeModule(Layer):
+    """Bridge Module used in EDCN
 
       Input shape
-        - A list of 3D tensor with shape: ``(batch_size,1,embedding_size)``. Its length is ``number of subnetworks``.
+        - A list of two 2D tensor with shape: ``(batch_size, units)``.
 
       Output shape
-        - 2D tensor with shape: ``(batch_size, embedding_size)``.
+        - 2D tensor with shape: ``(batch_size, units)``.
 
     Arguments
-       - **activation**: Activation function to use.
-       
-        - **l2_reg**: float between 0 and 1. L2 regularizer strength applied to the kernel weights matrix.
+        - **bridge_type**: The type of bridge interaction, one of 'pointwise_addition', 'hadamard_product', 'concatenation', 'attention_pooling'
 
-        - **seed**: A Python integer to use as random seed.
+        - **activation**: Activation function to use.
 
       References
         - [Enhancing Explicit and Implicit Feature Interactions via Information Sharing for Parallel Deep CTR Models.](https://dlp-kdd.github.io/assets/pdf/DLP-KDD_2021_paper_12.pdf)
 
     """
 
-    def __init__(self, bridge_type='attention_pooling', activation='relu', l2_reg=0, seed=1024, **kwargs):
+    def __init__(self, bridge_type='hadamard_product', activation='relu', **kwargs):
         self.bridge_type = bridge_type
         self.activation = activation
-        self.l2_reg = l2_reg
-        self.seed = seed
 
-        super(BridgeLayer, self).__init__(**kwargs)
+        super(BridgeModule, self).__init__(**kwargs)
 
     def build(self, input_shape):
         if not isinstance(input_shape, list) or len(input_shape) < 2:
             raise ValueError(
-                'A `AttentionPoolingLayer` layer should be called '
-                'on a list of at least 2 inputs')
+                'A `BridgeModule` layer should be called '
+                'on a list of 2 inputs')
 
         self.dnn_dim = int(input_shape[0][-1])
+        if self.bridge_type == "concatenation":
+            self.dense = Dense(self.dnn_dim, self.activation)
+        elif self.bridge_type == "attention_pooling":
+            self.dense_x = DNN([self.dnn_dim, self.dnn_dim], self.activation, output_activation='softmax')
+            self.dense_h = DNN([self.dnn_dim, self.dnn_dim], self.activation, output_activation='softmax')
 
-        self.dense = Dense(self.dnn_dim, self.activation)
-        self.dense_x = DNN([self.dnn_dim, self.dnn_dim], output_activation='softmax')
-        self.dense_h = DNN([self.dnn_dim, self.dnn_dim], output_activation='softmax')
-
-        super(BridgeLayer, self).build(input_shape)  # Be sure to call this somewhere!
+        super(BridgeModule, self).build(input_shape)  # Be sure to call this somewhere!
 
     def call(self, inputs, **kwargs):
         x, h = inputs
@@ -1543,7 +1540,7 @@ def call(self, inputs, **kwargs):
         elif self.bridge_type == "hadamard_product":
             return x * h
         elif self.bridge_type == "concatenation":
-            return self.dense(tf.concat(inputs, axis=-1))
+            return self.dense(tf.concat([x, h], axis=-1))
         elif self.bridge_type == "attention_pooling":
             a_x = self.dense_x(x)
             a_h = self.dense_h(h)
@@ -1553,12 +1550,10 @@ def compute_output_shape(self, input_shape):
         return (None, self.dnn_dim)
 
     def get_config(self):
-        base_config = super(BridgeLayer, self).get_config().copy()
+        base_config = super(BridgeModule, self).get_config().copy()
         config = {
             'bridge_type': self.bridge_type,
-            'l2_reg': self.l2_reg,
-            'activation': self.activation,
-            'seed': self.seed
+            'activation': self.activation
         }
         config.update(base_config)
         return config