Skip to content

_to_type

to_type

Compute the data types of tensors within a collection of data recursively.

This method can be used with Numpy data: 

data = {"x": np.ones((10,15), dtype="float32"), "y":[np.ones((4), dtype="int8"), np.ones((5, 3), dtype="double")],
    "z":{"key":np.ones((2,2), dtype="int64")}}
types = fe.backend.to_type(data)
# {'x': dtype('float32'), 'y': [dtype('int8'), dtype('float64')], 'z': {'key': dtype('int64')}}

This method can be used with TensorFlow tensors: 

data = {"x": tf.ones((10,15), dtype="float32"), "y":[tf.ones((4), dtype="int8"), tf.ones((5, 3), dtype="double")],
    "z":{"key":tf.ones((2,2), dtype="int64")}}
types = fe.backend.to_type(data)
# {'x': tf.float32, 'y': [tf.int8, tf.float64], 'z': {'key': tf.int64}}

This method can be used with PyTorch tensors: 

data = {"x": torch.ones((10,15), dtype=torch.float32), "y":[torch.ones((4), dtype=torch.int8), torch.ones((5, 3),
    dtype=torch.double)], "z":{"key":torch.ones((2,2), dtype=torch.long)}}
types = fe.backend.to_type(data)
# {'x': torch.float32, 'y': [torch.int8, torch.float64], 'z': {'key': torch.int64}}

Parameters:

Name Type Description Default
data Union[Collection, Tensor]

A tensor or possibly nested collection of tensors.

 required

Returns:

Type Description
Union[Collection, str]

A collection with the same structure as data, but with any tensors substituted for their dtypes.
Source code in fastestimator/fastestimator/backend/_to_type.py 
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
def to_type(data: Union[Collection, Tensor]) -> Union[Collection, str]:
    """Compute the data types of tensors within a collection of `data` recursively.

    This method can be used with Numpy data:
    ```python
    data = {"x": np.ones((10,15), dtype="float32"), "y":[np.ones((4), dtype="int8"), np.ones((5, 3), dtype="double")],
        "z":{"key":np.ones((2,2), dtype="int64")}}
    types = fe.backend.to_type(data)
    # {'x': dtype('float32'), 'y': [dtype('int8'), dtype('float64')], 'z': {'key': dtype('int64')}}
    ```

    This method can be used with TensorFlow tensors:
    ```python
    data = {"x": tf.ones((10,15), dtype="float32"), "y":[tf.ones((4), dtype="int8"), tf.ones((5, 3), dtype="double")],
        "z":{"key":tf.ones((2,2), dtype="int64")}}
    types = fe.backend.to_type(data)
    # {'x': tf.float32, 'y': [tf.int8, tf.float64], 'z': {'key': tf.int64}}
    ```

    This method can be used with PyTorch tensors:
    ```python
    data = {"x": torch.ones((10,15), dtype=torch.float32), "y":[torch.ones((4), dtype=torch.int8), torch.ones((5, 3),
        dtype=torch.double)], "z":{"key":torch.ones((2,2), dtype=torch.long)}}
    types = fe.backend.to_type(data)
    # {'x': torch.float32, 'y': [torch.int8, torch.float64], 'z': {'key': torch.int64}}
    ```

    Args:
        data: A tensor or possibly nested collection of tensors.

    Returns:
        A collection with the same structure as `data`, but with any tensors substituted for their dtypes.
    """
    if isinstance(data, dict):
        return {key: to_type(value) for (key, value) in data.items()}
    elif isinstance(data, list):
        return [to_type(val) for val in data]
    elif isinstance(data, tuple):
        return tuple([to_type(val) for val in data])
    elif isinstance(data, set):
        return set([to_type(val) for val in data])
    elif hasattr(data, "dtype"):
        return data.dtype
    else:
        return np.array(data).dtype