使用tf.data.dataset创建数据集时失败

Question

我尝试使用tf.data.Dataset.from_tensor_slices获取数据集，但我得到:

ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type float).

with tf.device('/cpu:0'):
train_data = tf.data.Dataset.from_tensor_slices((train['email'].values, train['label'].values))
valid_data = tf.data.Dataset.from_tensor_slices((validation['email'].values, validation['label'].values))

，train['email'].values是一个字符串数组

Answer 1

看起来你可能在你的输入数据集中混合了数据类型，如字符串和浮点值

工作样例代码

import tensorflow as tf
x_train = ['[e-mail]','[e-mail]','[e-mail]','[e-mail]']
y_train = ['0','1','0','1']
dataset = tf.data.Dataset.from_tensor_slices([x_train,y_train])
for element in dataset:
print(element)

tf.Tensor([b'[e-mail]' b'[e-mail]' b'[e-mail]' b'[e-mail]'], shape=(4,), dtype=string)
tf.Tensor([b'0' b'1' b'0' b'1'], shape=(4,), dtype=string)