python - key 错误 : "Invalid split train[:80%]. Available splits are: [' train']"

Question

这个问题在这里已经有了答案:

Split train data to train and validation by using tensorflow_datasets.load (TF 2.1) (2 个回答)

关闭 2年前。

在 Link: https://www.tensorflow.org/tutorials/images/transfer_learning

(raw_train, raw_validation, raw_test), metadata = tfds.load(
'cats_vs_dogs',
split=['train[:80%]', 'train[80%:90%]', 'train[90%:]'],
with_info=真，
as_supervised=真，
)

为什么会出错

KeyError: "无效的拆分火车[:80%]。可用拆分为:['train']"

Answer 1

试试这个代码我们可以在 TF 2 中将其拆分为:

(raw_train, raw_validation, raw_test), metadata = tfds.load(
'cats_vs_dogs',
split=[
   tfds.Split.TRAIN.subsplit(tfds.percent[:80]),
   tfds.Split.TRAIN.subsplit(tfds.percent[80:90]),
   tfds.Split.TRAIN.subsplit(tfds.percent[90:])
],
with_info=True,
as_supervised=True,
)