问tensorflow tf,argmax()函数怎么理解？

下面代码为什么得到这样的输出？
import tensorflow as tf
import numpy as np


test= np.array([[[1,2],[2,3]],[[4,5],[6,7]],[[3,9],[4,6]],[[6,3],[4,8]] ])
print("**********************")
print(test)
print(test.shape)
print("***********************")

x0 = tf.argmax(test,0)
y0 = tf.Session().run(x0)

x1 = tf.argmax(test, 1)
y1 = tf.Session().run(x1)

x2 = tf.argmax(test, 2)
y2 = tf.Session().run(x2)

print("xxxxxxxxxxxxx")
print(y0)
print("xxxxxxxxxxxxx")
print(y1)
print("xxxxxxxxxxxxx")
print(y2)