def build_graph(self, **kwargs):
user_sparse_inputs = {
uf['feat']: Input(shape=(1, ), dtype=tf.float32)
for uf in self.user_sparse_feature_columns
}
item_sparse_inputs = {
uf['feat']: Input(shape=(1, ), dtype=tf.float32)
for uf in self.item_sparse_feature_columns
}
model = Model(inputs=[user_sparse_inputs, item_sparse_inputs],
outputs=self.call(
[user_sparse_inputs, item_sparse_inputs]))
model.__setattr__("user_input", user_sparse_inputs)
model.__setattr__("item_input", item_sparse_inputs)
model.__setattr__("user_embed", self.user_dnn_out)
model.__setattr__("item_embed", self.item_dnn_out)
return model
# def model_test():
# user_features = [{'feat': 'user_id', 'feat_num': 100, 'feat_len': 1, 'embed_dim': 8}]
# item_features = [{'feat': 'item_id', 'feat_num': 100, 'feat_len': 1, 'embed_dim': 8}]
# model = Dssm(user_features, item_features)
# model.build_graph()
#
# model_test()
def build_graph(self, **kwargs):
user_sparse_inputs = {
uf['feat']: Input(shape=(1, ), dtype=tf.float32)
for uf in self.user_sparse_feature_columns
}
item_sparse_inputs = {
uf['feat']: Input(shape=(1, ), dtype=tf.float32)
for uf in self.item_sparse_feature_columns
}
hist_item_sparse_inputs = [{
uf['feat']: Input(shape=(1, ), dtype=tf.float32)
for uf in self.hist_item_sparse_feature_columns
} for i in range(self.hist_len)]
labels_inputs = Input(shape=(1, ), dtype=tf.int32)
model = Model(inputs=[
user_sparse_inputs, item_sparse_inputs, hist_item_sparse_inputs,
labels_inputs
],
outputs=self.call([
user_sparse_inputs, item_sparse_inputs,
hist_item_sparse_inputs, labels_inputs
]))
model.__setattr__("user_input", user_sparse_inputs)
model.__setattr__("item_input", item_sparse_inputs)
model.__setattr__("user_embeding", self.user_embedding)
model.__setattr__("item_embeding", self.item_embedding)
return model
def summary(self, **kwargs):
user_sparse_inputs = {uf['feat']: Input(shape=(1, ), dtype=tf.float32) for uf in
self.user_sparse_feature_columns}
item_sparse_inputs = {uf['feat']: Input(shape=(1, ), dtype=tf.float32) for uf in
self.item_sparse_feature_columns}
labels_inputs = Input(shape=(1,), dtype=tf.int32)
model = Model(inputs=[user_sparse_inputs, item_sparse_inputs, labels_inputs],
outputs=self.call([user_sparse_inputs, item_sparse_inputs,
labels_inputs]))
model.__setattr__("user_input", user_sparse_inputs)
model.__setattr__("item_input", item_sparse_inputs)
model.__setattr__("user_embeding", self.user_dnn_out)
model.__setattr__("item_embeding", self.item_dnn_out)
model.summary()
return model
