张量，如何收集索引列表的值？

t2 = tf.constant([[0, 11, 2, 3, 4],
                  [5, 61, 7, 8, 9],
                  [10, 11, 12, 13, 14],
                  [15, 16, 17, 18, 19]])
valid_mask = t2 <= 10
validIndex  =  tf.where(valid_mask)
print('validIndex',validIndex) # Expectation = Reality

print()

print('Final Output',tf.gather(t2,indices=validIndex)) # Hmm.. What ?

我的最终输出是

tf.Tensor(
[[[ 0 11  2  3  4]
  [ 0 11  2  3  4]]

 [[ 0 11  2  3  4]
  [10 11 12 13 14]]......

 [[10 11 12 13 14]
  [ 0 11  2  3  4]]], shape=(9, 2, 5), dtype=int32)

Expected

[0,2,3,4,5,7,8,9]

请帮助调试和纠正并解释发生了什么

Use tf.gather_nd or tf.boolean_mask:

import tensorflow as tf
t2 = tf.constant([[0, 11, 2, 3, 4],
                  [5, 61, 7, 8, 9],
                  [10, 11, 12, 13, 14],
                  [15, 16, 17, 18, 19]])
valid_mask = t2 <= 10
validIndex  =  tf.where(valid_mask)

print(tf.gather_nd(t2, indices=validIndex))
print(tf.boolean_mask(t2, valid_mask))

tf.Tensor([ 0  2  3  4  5  7  8  9 10], shape=(9,), dtype=int32)
tf.Tensor([ 0  2  3  4  5  7  8  9 10], shape=(9,), dtype=int32)

顺便说一句，根据您的情况，您的预期出局数应包括数字 10。