Method _create_model

modelzoo/dbmtl/train.py:148–245

(self)

            return tf.variable_scope(name,  partitioner=part, reuse=tf.AUTO_REUSE)

    def _create_model(self):
        TAG_COLUMN = ['tag_category_list', 'tag_brand_list']
        for key in TAG_COLUMN:
            self._feature[key] = tf.strings.split(self._feature[key], '|')

        with tf.variable_scope('dnn'):
            # dnn part
            key_dict={}
            with tf.variable_scope('input_layer',
                                partitioner=self._input_layer_partitioner,
                                reuse=tf.AUTO_REUSE):
                print('Adaptive emb = ', self._adaptive_emb, 'TF = ', self.tf)
                if self._adaptive_emb and not self.tf:
                    '''Adaptive Embedding Feature part 1 of 2'''
                    print('Adaptive Embedding Feature part 1 of 2')
                    adaptive_mask_tensors = {}
                    for col in HASH_INPUTS:
                        adaptive_mask_tensors[col] = tf.ones([args.batch_size],
                                                              tf.int32)
                    input_emb = tf.feature_column.input_layer(
                        self._feature,
                        self._feature_column,
                        adaptive_mask_tensors=adaptive_mask_tensors,
                        cols_to_output_tensors=key_dict)
                else:
                    input_emb = tf.feature_column.input_layer(
                        self._feature,
                        self._feature_column,
                        cols_to_output_tensors=key_dict)

            # Shared Layer
            with tf.variable_scope('bottom_dnn_'):
                shared_features = input_emb
                if self.bf16:
                    shared_features = tf.cast(shared_features, dtype=tf.bfloat16)

                for layer_id, num_hidden_units in enumerate(self._bottom_dnn):
                    with self._make_scope(f'bottom_dnn_layer_{layer_id}', self.bf16, self._dense_layer_partitioner) as shared_layer_scope:
                        shared_features = tf.layers.dense(shared_features,
                                                        units=num_hidden_units,
                                                        activation=None,
                                                        name=f'{shared_layer_scope.name}/dense')
                        shared_features = DNN_ACTIVATION(shared_features, shared_layer_scope.name)
                        self._add_layer_summary(shared_features, shared_layer_scope.name)

            # Specific Layer
            final_tower = []
            relations_outputs = {}
            for [tower_name, label_name, hidden_units] in self._towers:
                with tf.variable_scope(tower_name):
                    tower_input = shared_features

                specific_features = tower_input

                if self.bf16:
                    specific_features = tf.cast(specific_features, dtype=tf.bfloat16)

                for layer_id, num_hidden_units in enumerate(hidden_units):