mindspore.dataset.dataloader.default_convert

View Source On Gitee
mindspore.dataset.dataloader.default_convert(data)[source]

Default function for converting each NumPy array element into a mindspore.Tensor when batching is disabled in DataLoader.

  • If the input is a NumPy array and its dtype is not str, bytes or object, convert it into a mindspore.Tensor;

  • If the input is a NumPy numeric or boolean scalar, convert it into a mindspore.Tensor;

  • If the input is a Mapping, keep all the keys unchanged and convert the value of each key by calling this function recursively;

  • If the input is a Sequence, convert the element at each position by calling this function recursively;

  • Otherwise, leave it unchanged.

Parameters

data (Any) – A single data to be converted.

Returns

Any, the converted data.

Examples

>>> import numpy as np
>>> from mindspore.dataset.dataloader import default_convert
>>>
>>> default_convert(np.array([0, 1, 2]))
Tensor(shape=[3], dtype=Int64, value= [0, 1, 2])
>>>
>>> default_convert(np.int32(0))
Tensor(shape=[], dtype=Int32, value= 0)
>>>
>>> default_convert({"data": np.array([0, 1, 2])})
{'data': Tensor(shape=[3], dtype=Int64, value= [0, 1, 2])}
>>>
>>> default_convert([np.array([0, 1, 2]), np.array([3, 4, 5])])
[Tensor(shape=[3], dtype=Int64, value= [0, 1, 2]), Tensor(shape=[3], dtype=Int64, value= [3, 4, 5])]
>>>
>>> default_convert(np.array(["text"]))
array(['text'], dtype='<U4')