如何在tensorflow中将字符串转换为float32

Question

我是张量流的新用户。我有以下张量 A：

<tf.Tensor: shape=(3,), dtype=string, numpy=
array([b'1,2.0,32.29,42.1,16.0,0.60,46.0,0.0,26.0,0.0,-1.31,-0.9'],
  dtype=object)>

我想得到以下两个张量：

张量 B：将张量 A 转换为浮点张量并获取张量 A 的第一个元素：

<tf.Tensor: shape=(3,), dtype=float, numpy=
array([1],
  dtype=object)>

张量 C：将张量 A 转换为浮点张量并提取张量 B 的最后 4 个元素

<tf.Tensor: shape=(3,), dtype=float, numpy=
array([26.0,0.0,-1.31,-0.9],
  dtype=object)>

能否请您告诉我如何从张量 A 中获得张量 B 和张量 C？文档令人困惑。

Answer 1

temp_list = [1,2.0,32.29,42.1,16.0,0.60,46.0,0.0,26.0,0.0,-1.31,-0.9]
temp_tensor = tf.convert_to_tensor(temp_list)
temp_tensor[0]
temp_tensor[-4:]

您应该阅读https://www.tensorflow.org/guide/tensor 以进一步讨论。

更新一：

我试试

temp_str = "1,2.0,32.29,42.1,16.0,0.60,46.0,0.0,26.0,0.0,-1.31,-0.9"
temp_tensor = tf.convert_to_tensor(temp_str)
tf.cast(tf.strings.split(temp_tensor, sep=","),dtype=tf.float32)

它会引发错误：

UnimplementedError: Cast string to float is not supported [Op:Cast]

好像tensorflow不允许convert string to float。

我不确定这是否能满足您的需求：

tf.convert_to_tensor(np.array(tf.strings.split(temp_tensor, sep=",").numpy(),dtype="float32"))

更新二：

我找到了最好的答案：

temp_str = "1,2.0,32.29,42.1,16.0,0.60,46.0,0.0,26.0,0.0,-1.31,-0.9"
temp_tensor = tf.convert_to_tensor(temp_str)
tf.strings.to_number(tf.strings.split(temp_tensor, sep=","))