TypeError: must be real number, not Tensor

Question

老师您好，我在跑以下方程时遇到 “TypeError: must be real number, not Tensor” 的问题 我现在tensorflow的版本是1.13， python版本是3. 您知道该如何解决这个问题吗？
我的代码为：

init = tf.global_variables_initializer()
batch_size = 20
train_steps = 1000

with tf.Session() as sess:
    sess.run(init)
    for i in range(0, train_steps):
        batch_data, batch_lables = train_data.next_batch(batch_size)
        loss_val, accur_val, _ = sess.run([loss, accuracy, train_op], 
                                          feed_dict = {x: batch_data, y: batch_lables})
    
        if i % 500 == 0:
            print ('[Train] Step: %d, loss: %4.5f, acc: %4.5f' \
                % (i, loss_val, accuracy))

以下是报错信息

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
/var/folders/ht/s_lk_n7j245dpr0sk715tk500000gn/T/ipykernel_47765/1309685021.py in <module>
     12         if i % 500 == 0:
     13             print ('[Train] Step: %d, loss: %4.5f, acc: %4.5f' \
---> 14                 % (i, loss_val, accuracy))

TypeError: must be real number, not Tensor

Answer 1

正十七 2021-11-15 00:10:21

同学你好，

print的时候，你应该用你代码里sess.run输出的accur_val，accuracy是tensor，是图的一部分，是无法print的。而accur_val则是当前输入下accuracy的值。