Tensorflow Serving部署推荐模型
1、找到当前模型中定义的variables,并在此定义一个saver用于保存模型参数
def saveVariables(self):
variables_dict = {}
variables_dict[self.user_embedding.op.name] = self.user_embedding
variables_dict[self.item_embedding.op.name] = self.item_embedding
for v in self.reduce_dimension_layer.variables:
variables_dict[v.op.name] = v
self.saver = tf.train.Saver(variables_dict)
在模型的输入和输出的地方,尽量自行定义name,这样在之后的部署的时候会方便很多!
self.item_input = tf.placeholder("int32", [None, 1],name="gat_iteminput")
self.user_input = tf.placeholder("int32", [None, 1],name="gat_userinput")
...
self.prediction = tf.sigmoid(tf.reduce_sum(s