refactor feature columns

refactor feature columns 
- different features can use different `embedding_dim`
-  group-wise interaction is available by setting `group_name`

Author: 浅梦

.github/workflows/ci.yml

@@ -18,7 +18,7 @@ jobs:
     strategy:
       matrix:
         python-version: [3.5,3.6,3.7]
-        tf-version: [1.4.0,1.14.0,2.0.0]
+        tf-version: [1.4.0,1.15.0,2.0.0]
 
         exclude:
           - python-version: 3.7

README.md

@@ -18,7 +18,7 @@
 <!-- [![Gitter](https://badges.gitter.im/DeepCTR/community.svg)](https://gitter.im/DeepCTR/community?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge) -->
 
 
-DeepCTR is a **Easy-to-use**,**Modular** and **Extendible** package of deep-learning based CTR models along with lots of core components layers which can be used to build your own custom model easily.It is compatible with **tensorflow 1.4+ and 2.0+**.You can use any complex model with `model.fit()`and `model.predict()` .
+DeepCTR is a **Easy-to-use**,**Modular** and **Extendible** package of deep-learning based CTR models along with lots of core components layers which can be used to easily build custom models.It is compatible with **tensorflow 1.4+ and 2.0+**.You can use any complex model with `model.fit()`and `model.predict()` .
 
 Let's [**Get Started!**](https://deepctr-doc.readthedocs.io/en/latest/Quick-Start.html)([Chinese Introduction](https://zhuanlan.zhihu.com/p/53231955))
 

deepctr/__init__.py

@@ -1,6 +1,4 @@
-from . import layers
-from . import models
 from .utils import check_version
 
-__version__ = '0.6.3'
+__version__ = '0.7.0'
 check_version(__version__)

deepctr/inputs.py

@@ -9,7 +9,7 @@
 from .sequence import (AttentionSequencePoolingLayer, BiasEncoding, BiLSTM,
                        KMaxPooling, SequencePoolingLayer,WeightedSequenceLayer,
                        Transformer, DynamicGRU)
-from .utils import NoMask, Hash,Linear
+from .utils import NoMask, Hash,Linear,Add
 
 custom_objects = {'tf': tf,
                   'InnerProductLayer': InnerProductLayer,
@@ -38,5 +38,6 @@
                   'DynamicGRU': DynamicGRU,
                   'SENETLayer':SENETLayer,
                   'BilinearInteraction':BilinearInteraction,
-                  'WeightedSequenceLayer':WeightedSequenceLayer
+                  'WeightedSequenceLayer':WeightedSequenceLayer,
+                  'Add':Add
                   }

deepctr/layers/__init__.py

@@ -40,12 +40,12 @@ def build(self, input_shape):
         self.bn = tf.keras.layers.BatchNormalization(
             axis=self.axis, epsilon=self.epsilon, center=False, scale=False)
         self.alphas = self.add_weight(shape=(input_shape[-1],), initializer=Zeros(
-        ), dtype=tf.float32, name= 'dice_alpha')  # name='alpha_'+self.name
+        ), dtype=tf.float32, name='dice_alpha')  # name='alpha_'+self.name
         super(Dice, self).build(input_shape)  # Be sure to call this somewhere!
         self.uses_learning_phase = True
 
-    def call(self, inputs,training=None,**kwargs):
-        inputs_normed = self.bn(inputs,training=training)
+    def call(self, inputs, training=None, **kwargs):
+        inputs_normed = self.bn(inputs, training=training)
         # tf.layers.batch_normalization(
         # inputs, axis=self.axis, epsilon=self.epsilon, center=False, scale=False)
         x_p = tf.sigmoid(inputs_normed)
@@ -59,9 +59,10 @@ def get_config(self, ):
         base_config = super(Dice, self).get_config()
         return dict(list(base_config.items()) + list(config.items()))
 
+
 def activation_layer(activation):
     if activation == "dice" or activation == "Dice":
-        act_layer =  Dice()
+        act_layer = Dice()
     elif (isinstance(activation, str)) or (sys.version_info.major == 2 and isinstance(activation, (str, unicode))):
         act_layer = tf.keras.layers.Activation(activation)
     elif issubclass(activation, Layer):

deepctr/layers/activation.py

@@ -147,6 +147,8 @@ def __init__(self, hidden_units, activation='relu', l2_reg=0, dropout_rate=0, us
         super(DNN, self).__init__(**kwargs)
 
     def build(self, input_shape):
+        # if len(self.hidden_units) == 0:
+        #     raise ValueError("hidden_units is empty")
         input_size = input_shape[-1]
         hidden_units = [int(input_size)] + list(self.hidden_units)
         self.kernels = [self.add_weight(name='kernel' + str(i),

deepctr/layers/core.py

@@ -17,7 +17,7 @@
 from tensorflow.python.layers import utils
 
 from .activation import activation_layer
-from .utils import concat_fun,reduce_sum,softmax,reduce_mean
+from .utils import concat_func,reduce_sum,softmax,reduce_mean
 
 
 class AFMLayer(Layer):
@@ -843,7 +843,7 @@ def call(self, inputs, **kwargs):
             new_feature_list.append(
                 tf.reshape(new_result, (-1, int(pooling_result.shape[1]) * new_filters, embedding_size)))
 
-        new_features = concat_fun(new_feature_list, axis=1)
+        new_features = concat_func(new_feature_list, axis=1)
         return new_features
 
     def compute_output_shape(self, input_shape):
@@ -943,7 +943,7 @@ def call(self, inputs, training=None, **kwargs):
             raise ValueError(
                 "Unexpected inputs dimensions %d, expect to be 3 dimensions" % (K.ndim(inputs)))
 
-        inputs = concat_fun(inputs, axis=1)
+        inputs = concat_func(inputs, axis=1)
         Z = reduce_mean(inputs, axis=-1,)
 
         A_1 = tf.nn.relu(self.tensordot([Z, self.W_1]))
@@ -1030,7 +1030,7 @@ def call(self, inputs, **kwargs):
                  for v, w in zip(itertools.combinations(inputs, 2), self.W_list)]
         else:
             raise NotImplementedError
-        return concat_fun(p)
+        return concat_func(p)
 
     def compute_output_shape(self, input_shape):
         filed_size = len(input_shape)

deepctr/layers/interaction.py

@@ -178,7 +178,7 @@ def compute_mask(self, inputs, mask):
             return None
 
     def get_config(self, ):
-        config = {'supports_masking': self.supports_masking}
+        config = {'weight_normalization':self.weight_normalization,'supports_masking': self.supports_masking}
         base_config = super(WeightedSequenceLayer, self).get_config()
         return dict(list(base_config.items()) + list(config.items()))
 
@@ -460,7 +460,6 @@ def __init__(self, att_embedding_size=1, head_num=8, dropout_rate=0.0, use_posit
         self.supports_masking = supports_masking
 
     def build(self, input_shape):
-
         embedding_size = int(input_shape[0][-1])
         if self.num_units != embedding_size:
             raise ValueError(

deepctr/layers/sequence.py

@@ -116,7 +116,7 @@ def get_config(self, ):
         return dict(list(base_config.items()) + list(config.items()))
 
 
-def concat_fun(inputs, axis=-1,mask=False):
+def concat_func(inputs, axis=-1, mask=False):
     if not mask:
         inputs = list(map(NoMask(), inputs))
     if len(inputs) == 1:
@@ -188,3 +188,24 @@ def softmax(logits, dim=-1, name=None):
         return tf.nn.softmax(logits, dim=dim, name=name)
     else:
         return tf.nn.softmax(logits, axis=dim, name=name)
+
+
+class Add(tf.keras.layers.Layer):
+    def __init__(self, **kwargs):
+        super(Add, self).__init__(**kwargs)
+
+    def build(self, input_shape):
+        # Be sure to call this somewhere!
+        super(Add, self).build(input_shape)
+
+    def call(self, inputs, **kwargs):
+        if not isinstance(inputs,list):
+            return inputs
+        if len(inputs) == 1  :
+            return inputs[0]
+        if len(inputs) == 0:
+            return tf.constant([[0.0]])
+
+        return tf.keras.layers.add(inputs)
+def add_func(inputs):
+    return Add()(inputs)