输出正确且符合预期。trainhot[1] 是第二个(从 0 开始的索引)训练样本的标签,它是一维形状 (43,)。您可以使用下面的代码来更好地理解 tf.one_hot:
onehot = tf.one_hot([0, 0, 41, 42], 43, on_value=1, off_value=0)
with tf.Session() as sess:
onehot_v = sess.run(onehot)
print("v: ", onehot_v)
print("v shape: ", onehot_v.shape)
print("v[1] shape: ", onehot[1])
output:
v: [[1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0]
[1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1]]
v shape: (4, 43)
v[1] shape: Tensor("strided_slice:0", shape=(43,), dtype=int32)