获取参数0.0的类型<class'numpy.float32'=“”>无效，必须为字符串或张量。 （无法将float32转换为张量或操作。）

Question

我编写了一种Bi-lstm语言模型，并且在https://github.com/yangsaiyong/tf-adaptive-softmax-lstm-lm/blob/master/train_lm.py中使用了adaptive softmax

def train(train_data, vocabulary_size, args):   

    with tf.Session() as sess:

        model = BiRNNLanguageModel(vocabulary_size, args)
        sess.run(tf.global_variables_initializer())

        saver = tf.train.Saver()

        ckpt = tf.train.get_checkpoint_state(checkpoint_dir=args.model_path)
        if ckpt and ckpt.model_checkpoint_path:
            saver.restore(sess, ckpt.model_checkpoint_path)
            step = int(ckpt.model_checkpoint_path.rsplit('-', 1)[1])


        def train_step(batch_x):
            feed_dict = {model.x: batch_x, model.keep_prob: args.keep_prob}

            step, loss, _ = sess.run([model.global_steps, model.cost, model.train_op], feed_dict=feed_dict)

            if step % 100 == 0:
                print("step {0}: loss = {1}".format(step, loss))
                saver.save(sess, args.model_path+'bilstm-model', global_step=step)

        batches = batch_iter(train_data, args.batch_size, args.num_epochs)
        for batch_x in batches:
            train_step(batch_x)
            step = tf.train.global_step(sess, model.global_steps)

但是当我想存储模型时，错误就发生在下面

TypeError: Fetch argument 0.0 has invalid type <class 'numpy.float32'>, must be a string or Tensor. (Can not convert a float32 into a Tensor or Operation.)