python - 如何将 tensorflow 变量打印到小数点后两位？

Question

weights = tf.Variable(tf.truncated_normal([2,3]))
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print('Weights:')
    print(sess.run(weights))

    print("{0:2f}".format(sess.run(weights)))

第一个打印语句按预期工作。

Weights:
[[ 0.30919516  0.29567152  0.11157229] 
 [ 0.26642913 -0.21269836 -0.58886886]]

使用 str.format() 的第二个打印语句给出以下错误。

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-20-b0e4f93d01f8> in <module>()
  4     print('Weights:')
  5     print(sess.run(weights))
----> 6     print("{0:2f}".format(sess.run(weights)))

> TypeError: non-empty format string passed to object.__format__

除了下面的答案之外，我还发现 np.set_printoptions( precision=2) 也有效。

np.set_printoptions(precision=2)
weights = tf.Variable(tf.truncated_normal([2,3]))
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    print('Weights:')
    print(sess.run(weights)) 

Weights:
[[ 0.01 -0.42 -1.57]
 [-0.44  1.62  0.27]]

Answer 1

您尝试打印整个数组，并且 format 需要可以表示为 float 的单个值。尝试这样:

print(np.around(sess.run(weights), 2)
#[[ 0.31  0.30  0.11] 
# [ 0.27 -0.21 -0.59]]

此外，正确的格式是0:.2f